Analysis of Multiple Classifiers for Herbal Plant Recognition

Abstract

In this paper multiple classifiers are used for automatic plant recognition based on the shape of leaf features which are extracted from the leaf images using different herbal plants. Four different classifiers have been used namely SVM, KNN, Random Forest, and Logistic regression. Shape feature is considered as an important feature among all other features like- color, texture, vein structure, etc. as a leaf is available throughout the year and leaf shape contains more features to extract. Geometric shape features that are calculated as leaf’s length, width, area, perimeter, area of leaf enclosed in a rectangle, percentage of leaf in the rectangle, calculated pixels of leaf in four different quadrants are extracted during feature extraction. Among all cases, LR (Logistic regression) performed best in 7 cases while RF performed in 5 cases. SVM and LR model classifiers performed best with 95% accuracy for feature 6. KNN and RF model classifiers performed best with 90% and 93% accuracy respectively for feature 3.