Performance Evaluation of Statistical Classifiers for Shape Recognition with Morphological Shape Decomposition Method

In this study, 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 octagonal shape features are identified from 2D-shapes with the morphological shape decomposition technique. These features are reduced using principle component analysis. These reduced features are used to recognize the shapes with the above four classifiers. Experimental results show that the parametric classifier quadratic and non-parametric classifier Parzen gives the good recognition rate above 99% among other two classifiers for the morphological octagonal disk features.