Improving the classification accuracy of the method of the moments using aspect ratio normalization

Abstract

The method of moments has been used in several forms for shape recognition. Due to the dynamic range of the moments, high-order moment elements do not contribute significantly in the classification process and in some cases they reduce the classification accuracy. A normalization procedure, called aspect ratio normalization, which improves the classification accuracy, is discussed. The procedure is applied to a set of data to demonstrate its performance.<>