Fingerprint classification system

• US 5,572,597 A
• Filed: 06/06/1995
• Issued: 11/05/1996
• Est. Priority Date: 03/29/1994
• Status: Expired due to Fees

• Current Assignee: Leidos Innovations Technology, Inc.
• Original Assignee: Loral Corporation (Newell Brands Inc.)
• Inventors: Hilbert, Edward E., Chang, Chung-Fu
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): CHANG, JON CARLTON
• Application Number:
  US08/468,832
• Time in Patent Office: 518 Days
• Field of Search: 382/124, 382/125, 382/156, 382/157, 395/23
• US Class Current: 382/125
• CPC Class Codes: G06F 18/2321 using statistics or function optimisation, e.g. modelling of probability density functions
• CPC Class Codes: G06V 40/1359 Extracting features related to ridge properties; Determining the fingerprint type, e.g. whorl or loop
• CPC Class Codes: G06V 40/1365 Matching; Classification

Abstract

A technique for fingerprint classification and/or identification, in which a fingerprint is defined by areas containing patterns of ridges and valleys. At least one local pattern is determined using locations and characterizations of the fingerprint, which are indicated by a rapid change in direction of the ridges and valleys. The fingerprint is classified into types based upon the relative locations and characterizations of said local pattern(s). The fingerprint identification process can utilize minutiae location and angles as well as local pattern characterizations. Neural networks are utilized in determining the local patterns. The amount of data required to store data defining the fingerprints using the local pattern and/or minutiae techniques is significantly reduced.

Claims

1. A method for processing data representing an image of a fingerprint, comprising the steps of:

prescreening the data to determine locations of regions of interest in the fingerprint data;

for a determined location of a region of interest, extracting a set of feature vectors within an area that includes the determined location;

applying the extracted set of feature vectors to a plurality of input nodes of a multi-layer neural network, the plurality of input nodes comprising an input layer for inputting the extracted set of feature vectors, the multi-layer neural network further comprising a multi-node output layer, and at least one multi-node hidden layer, the multi-layer neural network operating in accordance with a set of weighting functions that are mapped onto said multi-layer neural network for weighting inputs to the plurality of input nodes, for weighting inputs to the nodes of the at least one hidden layer, and for weighting inputs to the nodes of the output layer;

from an output of the multi-layer neural network, determining a set of probabilities that the set of feature vectors represent individual ones of a plurality of predetermined local pattern types; and

determining in accordance with a rule based classification technique, from most probable ones of the local pattern types for the identified regions of interest, and also from locations of the local pattern types within the image of the fingerprint, a most probable fingerprint type that the fingerprint data represents;

wherein the step of applying includes a preliminary step of predetermining values of individual ones of the set of weights in accordance with a supervised, Bayesian model-based training sequence wherein features indicative of a predetermined set of fingerprint local pattern types are clustered within a feature space, wherein

the multi-layer neural network operates in accordance with the expression;

##EQU11## wherein the step of determining a set of probabilities operates in accordance with the expression;

##EQU12## in which expressions P_rNN (O_k /f) is the Bayesian a posteriori probability estimate for object O_k ;

a_k is the a priori probability;

F₁, F₂, F₃ are the saturated forms of F₁ (x)=x, F₂ (x)=e^x, and F₃ (x)=0 if x <

0 and F₃ (x)=x if x ≧

0;

b_j^k is the subgroup weightings;

c_i^JK, m_i^JK, and σ

_i^JK are respectively the relative weightings of the features f_i, the mean, and the standard deviation of the feature training values for feature f_i in subgroup j of the k-th object, and in which F₁, m_i^JK and σ

_i^JK are associated with the input layer of the neural network, F₂ and c_i^JK are associated with the at least one hidden layer of the neural network and F₃ and b_j^K are associated with the output layer of the neural network; and

wherein the function that applies weights to the plurality of input nodes of the multi-layer neural network is given by the expression;

##EQU13##

2. A method as set forth in claim 1, wherein the step of determining a most probable fingerprint type includes a step of considering also a location of minutiae within the fingerprint image data.

3. A method as set forth in claim 1, wherein the step of extracting a set of feature vectors extracts a plurality of feature vectors that are centered on the determined location.

4. A method as set forth in claim 1, wherein said pre-screening step includes the steps of performing ridge-valley processing and angle change detection processing.