US6529614B1
Expired / Lapsed
US
⏱️ 0.0 years remaining
Advanced miniature processing handware for ATR applications
Assignee: California Institute of Technology
•
Filing: 1999-08-04
•
Granted: 2003-03-04
•
Expires: 2019-08-04
Abstract
Abstract A Hybrid Optoelectronic Neural Object Recognition System (HONORS), is disclosed, comprising two major building blocks: (1) an advanced grayscale optical correlator (OC) and (2) a massively parallel three-dimensional neural-processor. The optical correlator, with its inherent advantages in parallel processing and shift invariance, is used for target of interest (TOI) detection and segmentation. The three-dimensional neural-processor, with its robust neural learning capability, is used for target classification and identification. The hybrid optoelectronic neural object recognition system, with its powerful combination of optical processing and neural networks, enables real-time, large frame, automatic target recognition (ATR).
Inventors
- 👤 Tien-Hsin Chao
- 👤 Taher Daud
- 👤 Anikumar Thakoor
Bibliographic Details
Application Number
—
Jurisdiction
US
Data Source
google_patents
Last Synchronized
2026-11-09 13:24
CPC Classifications
G; G06; G06E; G06E3/00; G06E3/001; G06E3/003; G06N; G06N3/00; G06N3/02; G06N3/04; G06N3/045; G06N3/0464; G06N3/06; G06N3/063; G06N3/08; G06N3/09; G06V; G06V10/00; G06V10/20; G06V10/255; G06V10/88
Claims Analysis
Total Claims: 26
Claim 1
Independent Claim
98 words
A hybrid optoelectronic neural object recognition system for automated fast target recognition of a target of interest, comprising: a grayscale optical correlator for optically detecting and segmenting a target of interest thereby producing a target of interest image, the grayscale optical correlator comprising a gray-scale spatial light modulator for use as a correlation filter; a neural-network trained with object templates for convolving with the target of interest image; and a neural three dimensional processor comprising an image-processing and target recognizing computer for identifying the target of interest by convolving the target of interest image with the neural network.
Claim 2
Dependent Claim
98 words
A hybrid optoelectronic neural object recognition system for automated fast target recognition of a target of interest, comprising: a grayscale optical correlator for optically detecting and segmenting a target of interest thereby producing a target of interest image, the grayscale optical correlator comprising a gray-scale spatial light modulator for use as a correlation filter; a neural-network trained with object templates for convolving with the target of interest image; and a neural three dimensional processor comprising an image-processing and target recognizing computer for identifying the target of interest by convolving the target of interest image with the neural network.
Claim 3
Dependent Claim
16 words
The system of claim 1 , wherein the neural three-dimensional processor comprises an application-specific integrated circuit.
Claim 4
Dependent Claim
17 words
The system of claim 2 , wherein the spatial light modulator is a high resolution correlation filter.
Claim 5
Dependent Claim
16 words
The system of claim 1 , herein the neural network is a mutilayer electronic neural network.
Claim 6
Dependent Claim
17 words
The system of claim 4 , wherein the neural network comprises a databank of eigenvector based images.
Claim 7
Dependent Claim
29 words
The system of claim 5 , wherein the neural three dimensional processor is further for converting the target of interest image into an eigenvector based target of interest image.
Claim 8
Dependent Claim
16 words
The system of claim 6 , wherein the eignenvector based images comprise 64 element feature vectors.
Claim 9
Dependent Claim
29 words
The system of claim 7 , wherein the neural three dimensional processor is further for converting the target of interest image into an eigenvector based 64 element feature vector.
Claim 10
Dependent Claim
18 words
The system of claim 8 , wherein the neural network comprises a databank created using an eigenimage computation.
Claim 11
Independent Claim
22 words
The system of clam 1 , wherein the system is for automated fast target recognition of a plurality of targets of interest.
Claim 12
Independent Claim
22 words
The system of clam 1 , wherein the system is for automated fast target recognition of a plurality of targets of interest.
Claim 13
Dependent Claim
32 words
The system of claim 10 , wherein the grayscale optical correlator is for optically detecting and segmenting the plurality of targets of interest, thereby producing a plurality of target of interest images.
Claim 14
Dependent Claim
20 words
The system of claim 11 , further comprising a frame buffer for storing the plurality of target of interest images.
Claim 15
Dependent Claim
40 words
The system of claim 12 , further comprising a column load interface processor chip for retrieving each of the target of interest images from the frame buffer and electrically passing each target of interest image to the neural three-dimensional processor.
Claim 16
Dependent Claim
34 words
The system of claim 13 , wherein the ground and airborne image-processing and target recognizing computer is for identifying each target of interest by convolving each target of interest image with the neural network.
Claim 17
Dependent Claim
17 words
The system of claim 1 , wherein the optical correlator further comprises an input spatial light modulator.
Claim 18
Dependent Claim
28 words
The system of claim 15 , wherein the optical correlator further comprises a collimator lens for focusing the target of interest image onto the input spatial light modulator.
Claim 19
Dependent Claim
19 words
The system of claim 16 , wherein the input spatial light modulator comprises a 640×480 pixel liquid crystal display.
Claim 20
Dependent Claim
26 words
The system of claim 17 , wherein the optical correlator further comprises a ferroelectric liquid crystal spatial light modulator for filtering the target of interest image.
Claim 21
Dependent Claim
24 words
The system of claim 18 , wherein the optical correlator further comprises a fourier transform lens for fourier transforming the target of interest image.
Claim 22
Dependent Claim
37 words
The system of claim 19 , wherein the optical correlator further comprises a polarizing cubic beam splitter for focusing the target of interest image from the fourier transform lens onto the ferroelectric liquid crystal spatial light modulator.
Claim 23
Dependent Claim
54 words
The system of claim 20 , wherein the optical correlator further comprises a half-wave plate inserted between the polarizing cubic beam splitter and the ferroelectric liquid crystal spatial light modulator to steer the polarization orientation of the target of interest image to ensure bipolar amplitude output from the ferroelectric liquid crystal spatial light modulator.
Claim 24
Dependent Claim
29 words
The system of claim 21 , wherein the optical correlator further comprises an output photodetector array for capturing and digitizing the target of interest image from the optical correlator.
Claim 25
Independent Claim
136 words
A method for automatic target recognition, comprising the steps of: computing a plurality of training images; downloading the plurality of training images into an optical correlator; capturing an image with a plurality of targets of interest; directing the captured image into the optical correlator; detecting the plurality of targets of interest in the captured image by segmenting the captured image using a gray scale spatial light modulator as a correlation filter; producing a plurality of target of interest images from the segmentations of the captured image; digitizing each target of interest image from the optical correlator; transforming each digitized target of interest image into an eignvector data element for each target; and classifying and identifying each target of interest by processing each corresponding eignvector data element using a neural network trained with an eigenvector-based target database.
Claim 26
Independent Claim
136 words
A method for automatic target recognition, comprising the steps of: computing a plurality of training images; downloading the plurality of training images into an optical correlator; capturing an image with a plurality of targets of interest; directing the captured image into the optical correlator; detecting the plurality of targets of interest in the captured image by segmenting the captured image using a gray scale spatial light modulator as a correlation filter; producing a plurality of target of interest images from the segmentations of the captured image; digitizing each target of interest image from the optical correlator; transforming each digitized target of interest image into an eignvector data element for each target; and classifying and identifying each target of interest by processing each corresponding eignvector data element using a neural network trained with an eigenvector-based target database.
References Cited (Backward Citations)
US4892370A
Cited
US4762397A
Cited
US5317651A
Cited
US5497253A
Cited
US5220642A
Cited
US5339090A
Cited
US5129040A
Cited
US5255362A
Cited
US5080464A
Cited
US4963725A
Cited
US5515189A
Cited
US5428711A
Cited
US5534704A
Cited
US5842191A
Cited
US5750994A
Cited
USPTO Annuity & Maintenance Windows
| Annuity Stage | Window Opens | Payment Deadline | Surcharge Period Until | Estimated Fee | Status |
|---|---|---|---|---|---|
| 3.5 Years (Year 4) | 2006-03-03 | 2006-09-02 | 2007-03-04 | $2000.00 | Window Closed / Lapsed |
| 7.5 Years (Year 8) | 2010-03-03 | 2010-09-02 | 2011-03-04 | $3760.00 | Window Closed / Lapsed |
| 11.5 Years (Year 12) | 2014-03-03 | 2014-09-02 | 2015-03-04 | $7700.00 | Window Closed / Lapsed |