Automatic Classification from Electropherograms

Hans Bruun Nielsen / Section for Numerical Analysis

Abstract

Electropherograms are spectres obtained by electrophoretic analysis. They are normally infected with a large background noise and a distortion. The report gives efficient algorithms for determining the background and distortion. Experiments show that the combined use of these algorithms leads to reliable identification of different varieties of wheat from their electropherograms.

Key words: Electrophoretic analysis, Spline functions, Huber estimation, Matching.

IMM Technical Report 1/95