Leaf identification using radial basis function neural networks and SSA based support vector machine.

Area of Science:

• Computer Vision
• Plant Science
• Machine Learning

Background:

• Accurate leaf identification is crucial for plant classification and ecological studies.
• Traditional methods often struggle with variations in leaf orientation and scale.
• Developing efficient and robust leaf identification algorithms is an ongoing research area.

Purpose of the Study:

• To propose an efficient and robust scheme for leaf type identification.
• To evaluate the performance of different classification techniques for leaf shape analysis.
• To demonstrate the superiority of an optimized Support Vector Machine (SVM) approach.

Main Methods:

• Leaf boundary points were fitted to a continuous contour using Radial Basis Function Neural Networks (RBFNN).
• Centroid calculation, distance normalization, and feature extraction algorithm time complexity were analyzed.
• Leaf shape features were classified using RBFNN, Support Vector Machines (SVM), and an SVM optimized with the Salp Swarm Algorithm (SSA).

Main Results:

• The proposed feature extraction scheme demonstrated independence from translation, rotation, and scaling.
• The Salp Swarm Algorithm (SSA) optimized SVM achieved significant improvements in classification accuracy compared to RBFNN and standard SVM.
• The study compared the efficiency of complex feature extraction algorithms with the proposed method.

Conclusions:

• The developed RBFNN-based contour fitting and feature extraction method is efficient and robust for leaf identification.
• The Salp Swarm Algorithm (SSA) offers a promising optimization technique for SVM in leaf classification tasks.
• This research contributes a more accurate and efficient approach to automated plant identification systems.