One-shot classification of 2-D leaf shapes using Distributed Hierarchical Graph Neuron (DHGN) scheme with k-NN classifier

This article presents a scalable approach for classifying plant leaves using the 2-dimensional shape feature. The proposed approach integrates a distributed recognition scheme called Distributed Hierarchical Graph Neuron (DHGN) for pattern recognition and k-nearest neighbor (k-NN) for pattern classification. With increasing amount of leaves data that can be captured using existing image gathering and processing technology, the ability for any particular classification scheme to produce high recall accuracy while adapting to large-scale dataset and data features is very important. The approach presented in this paper implements a one-shot learning mechanism within a distributed processing infrastructure, enabling large-scale data to be classified efficiently. The experimental results obtained through a series of classification tests indicate that the proposed scheme is able to produce high recall accuracy and large number of perfect recalls for a given plant leaves dataset. Furthermore, the results also indicate that the recognition procedure within the DHGN distributed scheme incurs low computational complexity and minimum processing time.