Research on the Classification of Statistically and Graphically Defined Patterns

Improved target detection and classification capability is the aim of the report. Two types of target clues or features are considered, spectral and spatial. The emphasis is placed on using automatic means for performing the detection with the classification being performed by a human operator. For...

Full description

Saved in:
Bibliographic Details
Main Authors: Gerhardt,Lester A, Drake,Kenneth W
Format: Report
Language:English
Subjects:
AUTOMATA
Computer Hardware
Cybernetics
DECISION MAKING
FEATURE EXTRACTION
GRAMMARS
INFORMATION THEORY
INFRARED RADIATION
INPUT OUTPUT DEVICES
LEARNING MACHINES
LOGIC CIRCUITS
PATTERN RECOGNITION
REMOTE SENSING
SEQUENTIAL MACHINES
SPECTRUM ANALYSIS
SPECTRUM ANALYZERS
SPECTRUM SIGNATURES
STATISTICAL ANALYSIS
STOCHASTIC PROCESSES
TARGET RECOGNITION
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Improved target detection and classification capability is the aim of the report. Two types of target clues or features are considered, spectral and spatial. The emphasis is placed on using automatic means for performing the detection with the classification being performed by a human operator. For the spectral clues, a statistical model was used to aid in the classification of spectral signatures. For the near IR and video region, a six dimensional space was sufficient for most problems. Results are given for a set of actual data covering seven different classes. Approaching the detection problem as a separate entity, optimized preprocessing spectral filters are developed. Binary filters which are proven to also be optimum are derived and an algorithm for finding them is given. (Author)