Apparatus for generating a feature matrix based on normalized out-class and in-class variation matrices

• US 5,161,204 A
• Filed: 06/04/1990
• Issued: 11/03/1992
• Est. Priority Date: 06/04/1990
• Status: Expired due to Fees

• Current Assignee: Datron Advanced Technologies Incorporated
• Original Assignee: Neuristics, Inc.
• Inventors: Narayanan, Venkatesh, Hutcheson, Timothy L., Srinivasan, Ramanujam, Or, Wilson, Ryan, Thomas W., Hunt, Bobby R., Mohan, Subramaniam, Wohlmut, Peter G.
• Primary Examiner(s): Moore, David K.
• Assistant Examiner(s): Cammarata, Michael
• Application Number:
  US07/533,113
• Time in Patent Office: 883 Days
• Field of Search: 382/43, 382/49, 382/54, 382/16, 382/14, 364/413.2, 364/827
• US Class Current: 382/157
• CPC Class Codes: G06F 18/21 Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
• CPC Class Codes: G06V 10/431 Frequency domain transformation; Autocorrelation
• CPC Class Codes: G06V 30/194 References adjustable by an adaptive method, e.g. learning
• CPC Class Codes: G06V 40/16 Human faces, e.g. facial parts, sketches or expressions
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation
• CPC Class Codes: G07C 9/25 using biometric data, e.g. fingerprints, iris scans or voice recognition

Abstract

A method and apparatus under software control for pattern recognition utilizes a neural network implementation to recognize two dimensional input images which are sufficiently similar to a database of previously stored two dimensional images. Images are first image processed and subjected to a Fourier transform which yields a power spectrum. An in-class to out-of-class study is performed on a typical collection of images in order to determine the most discriminatory regions of the Fourier transform. A feature vector consisting of the highest order (most discriminatory) magnitude information from the power spectrum of the Fourier transform of the image is formed. Feature vectors are input to a neural network having preferably two hidden layers, input dimensionality of the number of elements in the feature vector and output dimensionality of the number of data elements stored in the database. Unique identifier numbers are preferably stored along with the feature vector. Application of a query feature vector to the neural network will result in an output vector. The output vector is subjected to statistical analysis to determine if a sufficiently high confidence level exists to indicate that a successful identification has been made. Where a successful identification has occurred, the unique identifier number may be displayed.

Claims

1. A device for acquiring images, extracting image information, storing the image information, and comparing the image information with query image information comprising:

first means for acquiring images;

second means for processing said images into image information;

third means for determining the most distinctive aspects of said image information;

said third means including means for performing an In-Class to Out-of-Class study;

fourth means for forming and storing database feature vectors comprising the magnitudes of said most distinctive aspects of said image information; and

fifth means for querying said fourth means to determine whether a query feature vector is sufficiently similar to any of said database feature vectors,

wherein said means for performing an In-Class to Out-of-Class study includes;

means for generating an In-Class Variation Matrix;

means for generating an Out-Class Variation Matrix;

means for normalizing said In-Class Variation Matrix to form a normalized In-Class Variation Matrix;

means for normalizing said Out-Class Variation Matrix to form a normalized Out-Class Variation Matrix;

means for generating a feature matrix, each component of said feature matrix corresponding to a ratio between corresponding components of said normalized Out-Class Variation Matrix and said normalized In-Class Variation Matrix;

means for normalizing said feature matrix into a Normalized Feature Matrix;

means for partitioning said Normalized Feature Matrix into bricks;

means for prioritizing said bricks; and

means for creating a feature template vector whose elements correspond to a subset of said bricks.

2. The device of claim 1 wherein said fourth means includes neural network processor means adapted to store said data base feature vectors.

3. The device of claim 1 wherein said components of said feature matrix are set to zero where said corresponding components of either said normalized Out-Class Variation Matrix or said normalized In-Class Variation Matrix are zero.

4. The device of claim 2 wherein said neural network processor means is trained using backward error propagation.

5. The device of claim 4 wherein the structure of said neural network processor means includes at least one hidden layer.

6. The device of claim 4 wherein the structure of said neural network processor means includes two hidden layers.

7. The device of claim 6 including means for detecting outriders training of said neural network processor means.

8. The device of claim 7 including means for applying additional training cycles to feature vectors which result in detected outriders.

9. The device of claim 5 including means for detecting outriders during training of said neural network processor means.

10. The device of claim 9 including means for applying additional training cycles to feature vectors which result in detected outriders.