Unsupervised Learning Approach for Clustering Leaf Images

Abstract

Abstract With appropriate representation methods, the clustering techniques are found to be efficient with neural networks. The present work aims to propose various feature representation techniques for efficient clustering. The methods used for feature representation in this paper are, a method using random closed set, a method using edge information of input entity, a method that uses Huff transformation and a method that uses boundary moments. A comparative study of these representation methods for clustering the input objects using artificial neural networks, specifically Self-Organizing Map (SOM) is focused.