Investigating k-NN with an adequately chosen topology for date fruit classification

Abstract

The main objective of this research is to develop a new k-NN (k-Nearest Neighbors) classifier that offers improved accuracy over traditional k-NN classification methods, which typically use Euclidean metrics to define neighborhoods as circles around test points. In contrast, our approach utilizes covariance matrices to construct elliptical neighborhoods with rotations. The second aim is to apply the enhanced k-NN algorithm to classify date fruit species, which is challenging due to the diversity of varieties and cultivation practices. Classifying these fruits using external features is complex and labor-intensive; therefore, the proposed k-NN approach aims to improve efficiency in agricultural applications.