Plant Leaf Recognition using Geometric Features and Pearson Correlations

Abstract

Plant identification is an important task that is necessary for professionals like biologists, chemists, botanists, farmers, and nature hobbyists. The identification of plants from their leaves is a well-known strategy. In this paper, we present a novel set of features based on Pearson correlation coefficients, and we show the applicability of the proposed features for the classification of plant leaves. The foremost contribution in this paper is the use of the Pearson correlation coefficient computed from the leaf boundary pixels for analyzing shape similarity. The method has been tested on two well-known plant leaf datasets, Flavia and Swedish. The method shows the accuracy level of 95.16% on the Flavia dataset and of 97.0% on the Swedish dataset. The results corroborate the strength of our proposed feature set in comparison with other available methods.