International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.177
Volume 8 Issue I, Jan 2020- Available at www.ijraset.com
© IJRASET: All Rights are Reserved
282
Herbal Leaves Image Clustering via K-Means
Dr. A. Anushya
Assistant Professor, Department of Computer Applications
Abstract: India is a pervasive origin for the medicinal plants which fruitages traditional medicine. Modern research in computer
science is developing an automatic recognition system of plants by its leaf images. Features decide individual leaves. In this
research, features are extracted from herbal leaf images and clustered by Gray Level Concurrence Matrix and K-means
respectively. Both algorithms are executed on MATLAB and reached 86.96% accuracy.
Keywords: Gray Level Concurrence Matrix, K-means, Image clustering, herbal leaf images
I. INTRODUCTION
Traditional Medicine is time-tested and still caters to the health needs of the civilization and affords health care from side to side
prophylactic treatment and renovation [1]. All and sundry distinguish subsidy of herbal for our life. Herbals are instantaneous
remedy. Every kitchenette resides not only food stuffs, also includes some important herbals. Some of the herbals are used in our
diet regularly to enhance the flavour to various dishes. Apart from enriching the taste and aroma of the dish, they are used as
accepted cures for countless health welfares. These herbs can be cultivated in the backyard, terrace, and porch or in a pot also. This
research concentrates on these herbal which are used daily and keep in our kitchen cabinet. They are Basil leaves, Coriander, Mint,
Curry leaves, Purple fruited pea egg, Betel and Neem. At the beginning images are captured by using a mobile camera or a digital
camera. A complete database is assembled by packing every meticulous information of every image. Next, algorithm extracts
significant features from the images which are used for identical of the same in changed views. Identification of features consists of
staged filtering methodology. The images can be categorized to their respective species based on the clustering results.
The rest of paper is catalogued as follows: Section 2 bulletins the current research done in this field. Section 3 demarcates the data
source. Section 4 and Section 5 bounces the K-Means clustering for images and experimental results respectively. Section 6
concludes the paper and discusses the Future Works and Section gives the References.
II. RELATED WORKS
In this section, a meticulous review of the studies on herbal medicine image recognition and retrieval. Overall speaking, there is
pint-sized study on herbal image retrieval, while there are some on herbal image recognition. Therefore, we mainly introduce the
previous work on herbal image recognition, and simply review image retrieval in the computer vision community. Li [2] conducted
a research considered 5 herbal medicine categories using low-level features for medicine recognition, such as shape, color and
texture features. As the herbal medicine category becomes more, Tao et al. [3] found that color and shape features are not reliable
because many herbal medicine categories have the similar color and shape, while different medicine categories have different
textures. Therefore, they proposed to use various texture features from different aspects to describe the herbal images, and obtained
promising recognition precision on 18 herbal medicine categories. I.Kiruba Raji et.al., [4]examined various object detection
techniques for segmenting leaves based on color, shape and texture. Features like local adaptive mean color, evidence based color
model, color histogram techniques were used. Boundary structure model was used to detect the leaves based on boundary
descriptors of an image and Chan-Vese algorithm was used to segment the leaves from complex background. To extract leaves from
texture background, edge focusing algorithm was used. From our experimentation analysis, shape is the powerful characteristics of
segmenting leaf images and Chan-Vese algorithm provided better results compared to other techniques without affecting the leaf
colors, texture etc. Xin Sun et.al., [5] used the Convolutional Neural Network (CNN) for Chinese herbal medicine image recognition
and retrieval and practiced the softmax loss to optimize the recognition network; then for the retrieval problem, fine-tune the
recognition network by adding a triplet loss to search for the most similar medicine images. To appraise this method, a public
database of herbal medicine images constructed with cluttered backgrounds, which has in total 5523 images with 95 popular
Chinese medicine categories. Experimental results showed that the method can achieve the average recognition precision of 71%
and the average retrieval precision of 53% over all the 95 medicine categories, which are quite promising given the fact that the real
world images have multiple pieces of occluded herbal and cluttered backgrounds. Roopashree S.et.al., automated identification and
classification of Ayurveda leaf would be extraction of SIFT feature and implementing SVM classifier for better accuracy on using
Traditional Indian Medicine database. Also, portability could be achieved through different smartphone platforms.