Pyramided and optimized blurred shape model for plant leaf classification

Abstract

Plant leaf classification is a crucial task in the field of computer vision and pattern recognition, with various applications such as plant species identification and disease diagnosis. In this paper, the authors introduce the Pyramid Blurred Shape Model (PBSM) as a new descriptor for plant leaf classification. The PBSM extracts both shape and texture features from plant leaf images, which are combined to define a probability density function for leaf shape. Our experimental results show that the proposed PBSM achieves high classification accuracy, F1‐score, and precision‐recall results, demonstrating its effectiveness for leaf classification. However, extracting all available features from leaf images can lead to redundant and inessential features, which can degrade the classification performance and computational efficiency. To address this issue, the authors implement grey wolf optimization (GWO)‐based feature selection to identify the most informative features for classification. The final set of features is then classified using a list of selected classifiers, further enhancing the performance of the authors’ approach. The authors evaluate their proposed method on three publicly available datasets, namely the Middle European Woods (MEW), Swedish, and Flavia leaf datasets, and achieve high classification accuracies of 96.34%, 96.89%, and 92.41% for the Flavia, Swedish, and MEW leaf datasets, respectively. The authors’ approach outperforms state‐of‐the‐art descriptors in terms of classification accuracy and robustness, demonstrating its potential for real‐world applications. Overall, the authors’ proposed PBSM descriptor with feature selection provides a reliable and efficient solution for plant leaf classification. It can contribute to the development of automated plant species identification systems and disease diagnosis, thereby facilitating the conservation and protection of plant species.