Document Type

Thesis - Open Access

Award Date

1972

Degree Name

Master of Science (MS)

Department

Electrical Engineering

Abstract

The ability to make the correct decision is an ability that is perhaps desired by all human beings. Being able to make a correct decision is a function of the amount of information that is available to the decision maker. His ability to make a decision is enhanced by the use of machines specially designed for obtaining information or making decisions such as an electronic computer. Such is the case in the application of pattern recognition to a remote sensing problem. Pattern recognition is defined here as techniques which help distinguish between several classes such that the classification error is minimized. In the field of remote sensing, investigators have been given the task of classifying the data that appears on photographic imagery obtained from aerial flights. Much time would be saved if the data could be processed by automatic means. The primary purpose of this study is to determine the quantizer parameters which minimize the probability of error. The necessary conditions to minimize the probability of error for quantized data are derived. The same criterion will be applied to both continuous and quantized data. That is, the goal of the research is to determine the parameter conditions for the best quantizer and to determine if the error rate can approach that of the continuous case. A secondary objective relates to the number of required measurements. A set of curves with probability of error versus number of samples for varying degrees of accuracy are to be given. The set of curves are to provide a trade-off between accuracy and number of samples for a specified probability of error. As the number of samples approaches infinity, the probability of error will approach zero.

Library of Congress Subject Headings

Error-correcting codes (Information theory)
Remote sensing

Format

application/pdf

Number of Pages

98

Publisher

South Dakota State University

Share

COinS