Image analysis – Applications – Personnel identification
Reexamination Certificate
1999-03-02
2002-07-23
Mancuso, Joseph (Department: 2723)
Image analysis
Applications
Personnel identification
C382S209000
Reexamination Certificate
active
06424729
ABSTRACT:
FIELD OF THE INVENTION
The present invention generally relates to electro-optical systems for security verification, and in particular to an improved method and system using a joint transform correlator.
BACKGROUND
In a classical joint transform correlator system, a plurality of reference images are known, and a target image is unknown. The system is designed to determine whether the target image is the same as any of the reference images. A comparison is made by correlating the images optically; the correlation result is determined from detected cross-correlation peak values.
The target and reference images are displayed in a joint input scene. The scene is illuminated, resulting in an amplitude of light f(x,y) emanating from the joint input scene. The light is optically Fourier transformed, resulting in a complex amplitude F. The joint power spectrum, corresponding to |F|
2
, is detected. The joint power spectrum |F|
2
(not F) is then Fourier transformed, resulting in a cross-correlation pattern.
In general, if there is one target image and n reference images, the cross-correlation pattern has n+1 zeroth-order terms and n-choose-2 false alarm terms that are undesired. The cross-correlation pattern also has n desired cross-correlation terms. The power spectrum subtraction technique, known in the art, removes the undesired terms. A brief explanation can be found in “Improved Feature Extraction Using A Joint Wavelet Transform Correlator,” Boon Yi Soon, M. S. Alam, and M. A. Karim, Applied Optics, 37 (1998) pp. 821-827, and “Edge Detection Using Joint Transform Correlation,” Boon Yi Soon, M. S. Alam, and M. A. Karim, Proceedings of SPIE, 3073 (1997) pp. 343-353.
The spatial separation between the target and reference images needs to meet the Nyquist criterion in order for the terms in the cross-correlation pattern to be distinctive spatially from each other. When many reference images are used while still meeting the Nyquist criterion, the space-bandwidth product requirement needed to accommodate all images increases. Basically, the space-bandwidth product is the number of pixels necessary in the display device used for displaying the target and reference images. The most common display device is a spatial light modulator, or SLM. A high quality SLM is the most expensive component in the optical correlator. Therefore, when a large SLM is needed to display the joint input scene, the cost of implementation is very high.
Other techniques are available to perform optical security verification; one interesting proposal is presented in “Pattern Classification Using A Joint Transform Correlator Based Nearest Neighbor Classifier,” G. Lu and F. Yu, Optical Engineering 35 (1996) pp. 2162-2170. The primary goal of G. Lu et al. is to determine the correct character from a template of characters used as the reference images. In their correlation output, there is a location allocated for each reference image. They claim that their nearest neighbor classification scheme, which is based on the “maximum-win-all” algorithm, will verify the correct character by identifying the largest cross-correlation peak, which occurs at the location of the corresponding correct reference image in the correlation output.
However, the scheme of G. Lu et al. is inaccurate at times because the largest cross-correlation peak does not always correspond to the correct reference image, as illustrated in “Complementary-Reference Joint Transform Correlator,” Z. K. Chen, Y. Zhang, and G. Mu, Applied Optics, 33 (1994) pp. 7622-7626. When a plain white image correlates with a character, for example, the cross-correlation peak is higher than the cross-correlation peak of the character with itself Moreover, the cases when none of the reference images is the target image, and when two or more reference images are the target image, are not discussed.
Javidi et al., in U.S. Pat. No. 5,367,579, disclose a joint transform correlator of a target and reference image. The desired cross-correlation terms are optically separated from the unwanted terms by focusing the desired and undesired terms on different planes. However, diffuse light from the undesired terms still contaminates the desired signal.
Objects and Advantages
In view of the above, it is an object of the present invention to provide a system and method for fingerprint verification that does not require an expensive spatial light modulator. It is another object to provide an improved method for using cross-correlation peaks to determine whether the target image matches one of the reference images. It is a further object to provide a system and method that isolates cross-correlation peaks from optical contamination.
These objects yield a system and method having the following advantages: a target image such as a fingerprint can be checked for authenticity at reduced cost and with improved accuracy.
Summary
An improved system and method for optical fingerprint security verification is disclosed. The system comprises a hybrid electro-optics correlator comprising first and second spatial light modulators. The first spatial light modulator displays n reference images, where n is any positive number. The second spatial light modulator displays a target image. Because the target and reference images are separated, a large and expensive spatial light modulator is not needed.
The spatial light modulators are illuminated by monochromatic, polarized light. Light emanating from the first spatial light modulator and a portion of the light emanating from the second spatial light modulator are combined, jointly Fourier transformed, and detected by a first detector. The resulting joint power spectrum is stored in a computer.
Another portion of the light emanating from the second spatial light modulator is Fourier transformed and detected by a second detector. The resulting target-only power spectrum is also stored in the computer. The target-only and joint power spectra are detected simultaneously. A reference-only power spectrum corresponding to a Fourier transform of the reference images is stored in the computer as well.
The computer runs a program that preferably high-pass filters the target-only, reference-only, and joint power spectra. The program then obtains a modified power spectrum by subtracting the target-only and reference-only power spectra from the joint power spectrum. Because the high-pass filtering and subtraction are performed digitally rather than optically, a clean signal is produced.
The computer then displays an image corresponding to the modified power spectrum on the first spatial light modulator, and the first detector receives a cross-correlation pattern corresponding to a Fourier transform of the modified power spectrum. Cross-correlation peaks are identified, and maximum cross-correlation peak values A
i
are determined, for i=1 to n.
The values A
i
are subsequently processed using the following “uniqueness comparison scheme.” Each value A
i
is compared to a corresponding verification value A
i
. A tolerance, or range A is fixed by the user, and each value A
i
is tested to determine whether A
i
′−&Dgr;≦A
i
≦A
i
′+&Dgr; for one value of i. If A
i
falls within the range &Dgr; of A
i
′ for exactly one value of i, the target image is authenticated as being equivalent to the i
th
reference image. Otherwise, the target image is rejected. The uniqueness comparison scheme overcomes difficulties of the prior art, which rely only on the largest cross-correlation peak.
The system also includes a means for capturing the target image. A total-internally-reflecting prism, upon which a finger may be placed, is illuminated by monochromatic light. Light reflected from the interface between the prism and the finger carries an image of a fingerprint; the reflected light is detected by a third detector stored by the computer.
REFERENCES:
patent: 5224173 (1993-06-01), Kuhns et al.
patent: 5367579 (1994-11-01), Javidi et al.
patent: 5699449 (1997-12-01), Javidi
patent: 5815598 (1998-09-01), Hara et
Bali Vikkram
Lumen Intellectual Property Services Inc.
Mancuso Joseph
LandOfFree
Optical fingerprint security verification using separate... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Optical fingerprint security verification using separate..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Optical fingerprint security verification using separate... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2868874