Author : Juby George 1
Date of Publication :27th April 2018
Abstract: Identification of leaves from digital images using various automatic pattern recognition algorithms results in performance degradations. Here, various leaf features are been extracted and exposed to Multi-layer Perceptron and Support Vector Machine. The leaf images are taken from the Columbia Dataset and that are preprocessed to get the region of interest (ROI). The shape, color and vein features are then extracted from the selected ROI. The prominent features are then found out by using Principal Component Analysis (PCA). The reduced feature sets are supplied to these algorithms for identification. Total 150 samples were taken from this dataset spreads over 10 different leaf species. Multilayer Perceptron and Support vector machine are trained with 104 leaf images and are been tested and validated using 23 leaf images each. A comparison is made between the performances of these two learning methods and found that the leaf recognition accuracy of the support vector machine is better than that of Multilayer Perceptron algorithm
Reference :