Optical pattern recognition architecture implementing the mean-square error correlation algorithm

• US 5,060,282 A
• Filed: 03/26/1990
• Issued: 10/22/1991
• Est. Priority Date: 03/26/1990
• Status: Expired due to Fees

• Current Assignee: United States Department of Energy (Government of the United States of America)
• Original Assignee: United States Department of Energy (Government of the United States of America)
• Inventors: Molley, Perry A.
• Primary Examiner(s): Boudreau, Leo H.
• Assistant Examiner(s): Fox, David
• Application Number:
  US07/499,120
• Time in Patent Office: 575 Days
• Field of Search: 350/162.11, 350/162.12, 350/162.13, 350/162.14, 382/31, 382/42
• US Class Current: 382/211
• CPC Class Codes: G06E 3/005 using electro-optical or opto-electronic means
• CPC Class Codes: G06V 10/88 Image or video recognition using optical means, e.g. reference filters, holographic masks, frequency domain filters or spatial domain filters

Abstract

An optical architecture implementing the mean-square error correlation algorithm,

MSE=Σ[I-R].sup.2

for discriminating the presence of a reference image R in an input image scene I by computing the mean-square-error between a time-varying reference image signal s₁ (t) and a time-varying input image signal s₂ (t) includes a laser diode light source which is temporally modulated by a double-sideband suppressed-carrier source modulation signal I₁ (t) having the form

I.sub.1 (t)=A.sub.1 [1+√2m.sub.1 s.sub.1 (t)cos (2πf.sub.o t)]

and the modulated light output from the laser diode source is diffracted by an acousto-optic deflector. The resultant intensity of the +1 diffracted order from the acousto-optic device is given by:

I.sub.2 (t)=A.sub.2 [+2m.sub.2.sup.2 s.sub.2.sup.2 (t)-2√2m.sub.2

(t) cos (2πf_o t]

The time integration of the two signals I₁ (t) and I₂ (t) on the CCD deflector plane produces the result R(τ) of the mean-square error having the form:

R(τ)=A.sub.1 A.sub.2 {[T]+[2m.sub.2.sup.2·∫s.sub.2.sup.2

(t-τ)dt]-[2m₁ m₂ cos (2τf_o τ)·∫s₁ (t)s₂ (t-τ)dt]}

where:

s₁ (t) is the signal input to the diode modulation source:

s₂ (t) is the signal input to the AOD modulation source;

A₁ is the light intensity;

A₂ is the diffraction efficiency;

m₁ and m₂ are constants that determine the signal-to-bias ratio;

f_o is the frequency offset between the oscillator at f_c and the modulation at f_c +f_o ; and

a_o and a₁ are constant chosen to bias the diode source and the acousto-optic deflector into their respective linear operating regions so that the diode source exhibits a linear intensity characteristic and the AOD exhibits a linear amplitude characteristic.

Claims

1. An optical pattern recognition architecture implementing the mean-square-error algorithm, MSE=Σ

[I-R]², for discriminating a reference pattern R in an input image I, comprising;

light source means for outputting modulated light in accordance with a light source modulation signal applied to a light source modulation signal input thereof;

optical deflector means for diffracting light incident at an aperture thereof in accordance with an optical deflector modulation signal applied to a signal input thereof for producing a spatial distribution of said incident light;

means for focusing modulated light output by said light source means onto the aperture of said optical deflector means;

means for generating a double-sideband suppressed-carrier amplitude modulated light source modulation signal I₁ (t) as the product of a time-varying reference image signal s₁ (t) applied thereto and a frequency offset carrier signal f_o and for applying said light source modulation signal I₁ (t) to said light source modulation signal input of said light source means for causing said light source means to be temporally modulated in accordance therewith, said light source modulation I₁ (t) taking the form;

space="preserve" listing-type="equation">I.sub.1 (t)=A.sub.1 [1+√

2m.sub.1 s.sub.1 (t) cos (2π

f.sub.o t)];

means for generating a double-sideband suppressed-carrier amplitude modulated optical deflector modulation signal I₂ (t) as the product of a time-varying input image signal s₂ (t) input thereto and a reference carrier f_c and an offset frequency carrier f_o, and for applying said optical deflector modulation signal I₂ (t) to said optical deflector modulation signal input of said optical deflector means for causing said modulated light output from said light source means to be diffracted into a time-delayed spatial modulation in accordance therewith, said optical deflector modulation I₂ (t) taking the form;

space="preserve" listing-type="equation">I.sub.2 (t)=A.sub.2 [1+2m.sub.2.sup.2 s.sub.2.sup.2 (t)-2√

2m.sub.2 s.sub.2 (t) cos (2π

f.sub.o t)];

integrating light detector means having a detection plane for detecting and electronically integrating light incident on said detector plane and for outputting a mean-square-error correlation signal R(t) in correspondence therewith, said mean-square-error correlation signal R(τ

) taking the form;

##EQU7## such that by adjusting the m₁ and m₂ for the input modulation, m₁ is equal to 2 m₂, and a zero value of R(τ

) represents a match correlation between said input image I and said reference pattern R; and

means for imaging the spatially distributed light diffracted by said optical deflector means onto the detector plane of said integrating light detector means;

where;

s₁ (t) is the signal input to the light source modulation means;

s₂ (t) is the signal input to the optical deflector modulation means;

A₁ is the light intensity;

A₂ is the diffraction efficiency;

m₁ and m₂ are constants that determine the signal-to-bias ratio;

f_o is the frequency offset between the reference oscillator at f_c and the DSB-SC modulation at f_c +f_o ; and

a_o and a₁ are constants chosen to bias the light source means and the optical deflector into their respective linear operating regions so that the light source means exhibits a linear intensity characteristic and the optical deflector means exhibits a linear amplitude characteristic.

2. An optical pattern recognition architecture according to claim 1, wherein the light source means is at least one laser diode.

3. An optical pattern recognition architecture according to claim 1, wherein the optical deflector means is an acousto-optic deflector.

4. An optical pattern recognition architecture according to claim 1, wherein the means for generating the optical deflector modulation signal I₂ (t) alters the phase of the offset frequency f_o by 180°

.

5. An optical pattern recognition architecture according to claim 1, wherein the image signal s₂ (t) is composed of a first D. C. component D₁ and a signal component g(t), and wherein the means for generating the optical deflector modulation signal I₂ (t) subtracts said D. C. component D₁ from said image signal s₂ (t) and adds a second D. C. component D₂ to said image signal s₂ (t), where

6. An optical pattern recognition architecture according to claim 1, wherein the integrating light detector means is a CCD photodetector array operated in a time-delay-and-integrate mode.

7. An optical pattern recognition architecture according to claim 6, further comprising decision logic means provided at a correlation signal output of the CCD photodetector for determining whether the value of said mean-square error correlation signal R(τ

) is below a predetermined threshold level corresponding to a desired correlation result between said input image I and said reference image R.

8. An optical pattern recognition architecture according to claim 6, wherein the CCD photodetector is provided with anti-blooming gate means incorporated with photodetector elements of said CCD photodetector for preventing saturated detector wells from interfering with data in adjacent detector wells.

9. An optical pattern recognition architecture according to claim 6, wherein the value of the relative light source intensity A₁ is set such that the charge level produced in said CCD photodetector, when said mean-square-error correlation signal R(τ

) assumes a value corresponding to a desired correlation result between said input image I and said reference image R, is within the dynamic range of said CCD photodector.

10. An optical pattern recognition architecture according to claim 6, wherein the CCD photodetector is provided with a sufficient number of pixels for detecting variations in the carrier term, cos [2π

f_o τ

], in said mean-square-error correlation signal R(τ

).

11. An optical pattern recognition architecture according to claim 1, wherein the means for imaging the light diffracted by the optical deflector means onto the detector plane of the light detector means includes a field flattener lens provided in front of the detector plane of the integrating light detector means.

12. An optical pattern recognition architecture according to claim 1, further comprising D. C. stop means for blocking undiffracted light from the optical deflector means from reaching the detector plane of the integrating light detector means.

13. An optical pattern recognition architecture according to claim 1, wherein the input signals s₁ and s₂ are bandlimited to a bandwidth B such that |S(f)|=0 for |f|>

B, and have unit variance;

the light source intensity and the modulation of the optical deflector have respective variances of m₁² and m₂², and the light source with modulation depth m is limited to ±

m/m₁ standard deviations of the input signal s₁.