Abstract

This paper is one of the early attempts for the recognition of Nandinagari handwritten characters. Good literature for finding specific types of key interest points using single approach is available for manuscripts. However, careful analysis indicate that a combinatorial approach is needed to be used in a collaborative manner for achieving good accuracy. On a variant data set of over 1000 Handwritten Nandinagari characters having different size, rotation, translation and image format, we subject them to an approach at every stage where their recognition is effective. In the first stage, the key interest points on the images which are invariant to Scale, rotation, translation, illumination and occlusion are identified by choosing Scale Invariant Feature Transform method. These points are then used to compute dissimilarity value with respect to every other image. Subsequently we subject these to Hierarchical Agglomerative cluster analysis for classification without supervision. Finally, for a query image, the same steps are followed and cluster mapping is analyzed. The result shows over 99% recognition, thus achieving a robust and accurate manuscript character recognition system.