ROBUST RECOGNITION AND CLASSIFICATION OF HERBAL LEAVES

Prabhakar Poudel, Shyamdew Kumar, Vishal Shaji Philip, Pawan Kishore, Roopashree S

Abstract: Our ecosystem has a large variety of plants and they are an essence for other forms of life to flourish. Identifying plants can be simpler if we just identify the leaf of a plant, because each of them has peculiar properties. Plants are often important ingredient for Ayurvedic medicine and also several other modern forms of medication. In this paper we discuss the use of optical method, for the classification of herbs using image processing techniques. Steps needed to execute in an optical method for the herbal leaf recognition include extraction of diverse features, identification of pattern and categorization in sequence. We propose an automation system for the herbal leaf detection to increase consistency, lessen the required time and ease the tagging procedure. A proposed technique also considerably reduces the possibility for human faults, as much of the work is done by the system. This paper hinge on for usage of Support Vector Machine (SVM) for the purpose of classification and Scale Invariant Feature Transform (SIFT) technique to extract features. SIFT features are proved to be invariant to affine transformations, noise and change in illumination. A captured leaf image using a digital camera is pre-processed which gives a binary image as output. The pre-processed image is taken up for extracting essential features, which ultimately is used for the purpose of classification. The projected scheme is suitable to use and effective in terms of cost for the researcher, botanist, and others to discover herbal plants, with high accuracy and effectiveness.

Keywords: Herbs, Ayurveda, Feature Extraction, Classifier, SIFT, SVM, Image Processing.

DOI: https://doi.org/10.15623/ijret.2016.0516031