Shape Classification with Statistical Classifiers using Morphological Shape Representation Features

This paper compares the performance of four statistical classifiers, namely linear discriminant classifier, quadratic discriminant classifier, k-Nearest Neighborhood classifier, and parzen classifier are considered for recognition of 2D-shapes. The two features from morphological skeletons and four features from morphological shape decomposition are identified from 2D-shapes. These features are reduced using Principle Component Analysis (PCA). These reduced features are used to recognize the shapes with the above four statistical classifiers. Experimental results show that the non-parametric classifier 3-NNC gives the good recognition rate 100% among other classifiers for the combined features of morphological octagonal disk and skeletons.