Leaf Recognition System using Image Processing and Support Vector Machine

Abstract

Identifying plants manually will require a demanding procedure. It includes knowledgeable individuals and consumes time.  Plant species are of great number, and unidentified plants increase, with this,  a plant recognition system is needed. The researcher attempted to develop an image processing-based system that recognizes automatically leaf parameters and uses extracted parameters for plant classification. This study utilized Industrial Revolution 4.0 by applying machine learning in classification. The study is very helpful in new plant discovery and plant taxonomy. Three plant species were used in this study, namely, Morus Alba, Annona Muricata, and ManilkaraZapota. Using Matlab software, the RGB Histogram of leaf image was taken and used to threshold and segment the leaf edges. Raw image data were denoised before leaf feature extraction to eliminate unwanted data. The result of segmentation provides extracted morphological features, and these are the leaf's area, the leaf's major and minor axes,  eccentricity, centroid, and leaf perimeter.  A ratio of 70:30 training dataset and the testing dataset was used to trained raw image data into different machine learning classifiers.  The classifier with the highest accuracy during the learning period was selected as the classifier model. The study shows that the use of  Support Vector Machine (SVM), a technique in the classification of extracted morphological features and shapes was able to obtain the highest accuracy.  Support Vector Machine model was able to classify test datasets containing random Morus Alba, Annona Muricata, and ManilkaraZapota leaves significantly with an overall accuracy of 100%, precision, and recall of 100%. The use of image processing is an effective technique in obtaining leaf parameters and in classifying them.