US 6,278,799 B1 · Hierarchical data matrix pattern recognition system

Abstract

The present invention relates to a hierarchical artificial neural network (HANN) for automating the recognition and identification of patterns in data matrices. It has particular, although not exclusive, application to the identification of severe storm events (SSEs) from spatial precipitation patterns, derived from conventional volumetric radar imagery. To identify characteristic features a data matrix, the data matrix is processed with a self organizing network to produce a self organizing feature space mapping. The self organizing feature space mapping is processed to produce a density characterization of the feature space mapping. The self organizing network is preferably completely unsupervised. It may, under some circumstances include a supervised layer, but it must include at least an unsupervised component for the purposes of the invention. The “self organizing feature space” is intended to include any map with the self organizing characteristics of the Kohonen Self Organizing Feature Map. The frequency vector of a CAPPI image that has been derived is a data abstraction that can be displayed directly for examination. In preferred embodiments, it is presented to a classification network, e. g. the standard CPN network, for classifying the density vector representation of the three dimensional data and displaying a representation of classified features in the three dimensional data. A novel methodology is preferably used for incorporating vigilance and conscience mechanisms in the forward counterpropagation network during training.

Claims25

Dependent claims

A method of processing plural object data matrices to identify characteristic features therein, said method comprising:

A producing a trained self organizing feature space mapping by:
i) providing a set of training data matrices containing the characteristic features to be identified;
ii) fragmenting each of the training data matrices into plural elements;
and
iii) processing each element of each training data matrix with a single self organizing network to produce a single trained self organizing feature space mapping (SOFM);
and
B identifying the characteristic features in each of the object data matrices by:
iv) fragmenting one of the object data matrices into plural object matrix elements;
v) processing the object matrix elements with the single SOFM to produce a feature characterization of the object data matrix;
vi) processing the feature characterization to produce a density characterization of characteristic features in the object data matrix;
and
vii) repeating steps iv), v) and vi) for each of the object data matrices.

9 more elements

A system for processing plural object data matrices to identify characteristic features therein, said system comprising:

A means for producing a trained self organizing feature space mapping, including:
i) a set of training data matrices containing the characteristic features to be identified, each training data matrix fragmented into plural elements;
and
ii) a single self organizing network for processing each element of each training data matrix to produce a single trained self organizing feature space mapping (SOFM);
B means for identifying the characteristic features in each of the object data matrices including:
iii) means for fragmenting each of the object data matrices into plural object matrix elements;
iv) means for processing the object matrix elements with the single SOFM to produce a feature characterization of each object data matrix;
and
v) density map processing means for processing each feature characterization to produce a density characterization of the characteristic features in the object data matrices.

6 more elements

Dependencies are inferred from the claim text.

Hierarchical data matrix pattern recognition system

Life of this patent

Hierarchical data matrix pattern recognition system

Life of this patent

Abstract

Claims25