A nonmetric approach to measurement: an application to pattern representation

A nonmetric-nonstatistical method for representing an N-point configuration is presented as an alternative to metric methods such as using Cartesian coordinates. The advantages of a nonmetric basis include the ability to use sensors (measurements) of poor rather than high quality by substituting novel computational algorithms to offset sensor inaccuracy. An outline of this approach to measurement is presented and supported by computer simulation results. The simulations illustrate a nonnumeric encoding and recovery of geometric pattern information and also furnish interesting insights into human vision