Identification of Lakes by Closed Planar Curve Analysis

Abstract

The experiments to identify major Japanese lakes from the remote sensing pictures are carried out by applying the technique of closed planar curve analysis using the argument function.
As the argument function is insensitive to most of spatial operations including translation, magnification and reduction, it is widely used in the field of pattern re-cognition of characters, chromosomes, and so on.
The contours of 9 lakes in Japan from the LANDSAT-1 MSS (Multi-Spectral Scanner) pictures, and those of 21 lakes from a map as model patterns, are extracted and transformed into the argument functions.
In the first experiment, the normalized argument functions are directly compared, and in the second one, the power spectra of the argument functions are used, to identify the lakes with good result.
It is concluded that the use of closed planar curve is a very effective method for the identification of lakes.