2012 International Conference on Advanced Computer Science and Information Systems (ICACSIS) 1-2 Des 2012. pp. 255-259.
Combination of Morphological, Local Binary Pattern Variance and Color Moments Features for Indonesian Medicinal Plants Identification Yeni Herdiyeni, Mayanda Mega Santoni Department of Computer Science Faculty of Mathematic and Natural Science, Bogor Agricultural University e-mail:
[email protected];
[email protected]
Abstract— We propose a new method for Indonesian medicinal plants identification using combination of some leaf features, i.e. texture, shape, and color. Local Binary Pattern Variance (LBPV) is used to extract leaf texture, morphological feature is used to extract leaf shape, and color moment is used to extract leaf color distribution. In the experiment we used 51 species of Indonesian medicinal plants and each species consists of 48 images, so the total images used in this research are 2,448 images. Combination of leaf feature is done using Product Decision Rule (PDR) and classification of medicinal plants is done using Probabilistic Neural Network (PNN). The experimental results show that the combination of the morphological, LBPV, and color moments features can improve the accuracy of medicinal plants identification. This research is important to enhance utilization of Indonesian medicinal plants. Index Terms—color moments, local binary pattern variance, Indonesian medicinal plant identification, morphological, probabilistic neural network
I
I. INTRODUCTION
ndonesia is a country of mega biodiversity. Indonesian Science Board (Lembaga Ilmu Pengetahuan Indonesia/LIPI) states that Indonesia is home to 30,000 out of 40,000 medicinal herbal plants in the world. Some plants for medicinal use has been documented by researchers, students, and practitioners through the exploration of the various regions in Indonesia, either through surveys of potential plant diversity and ethno botany studies [1]. In 2001 Department of Forest Resources Conservation and Ecotourism, Faculty of Forestry, Bogor Agricultural University (Institut Pertanian Bogor – IPB) has been collected and recorded more than 2,039 species of Manuscript received September 21, 2012. Y. Herdiyeni and N.K.S.Wahyuni are with the Department of Computer Science, Faculty of Mathematics and Natural Sciences, Bogor Agricultural University. Bogor 16680 (e-mail: yeni.herdiyeni@ ipb.ac.id, megasantoni@ gmail.com).
medicinal plants from forest in Indonesia [2]. According to Groombridge and Jenkins [3], the percentage of Indonesian medicinal plant used has only been 4.4%. Currently, to identify medicinal plants specimens should be brought in herbarium. Identification of herbarium specimens was limited and time-consuming with few taxonomists able to identify them. Identification of medicinal plants is not easy because of their large number (both of medicinal and not medicinal). On the other hand, in the mid of 2012, it is estimated that two third of Indonesia’s digital consumer or 67% of the market, own smartphone. It is shows that mobile technology can be used as a tool to disseminate and identify of medicinal plant information to the public automatically. To overcome these problems, in this research we propose a mobile application for medicinal plant identification based on image to facilitate the user identifying and finding information about medicinal plants using leaf images automatically. A number of researches have been explored to extract or measure leaf features [4]-[11]. Wu et.al [4] uses Probabilistic Neural Network (PNN) for leaf recognition. Gu et al. tried leaf recognition using skeleton segmentation by wavelet transform and Gaussian interpolation [12]. This research develops a mobile application for identifying Indonesian medicinal plant using combination of leaf features. The application runs on Android mobile operating system. Our main improvements are on combination of leaf feature and classifier combination. Leaf features consists of leaf texture, leaf shape and leaf color distribution. Local Binary Pattern Variance (LBPV) is used to extract leaf texture, morphological feature is used to extract leaf shape, and color moment is used to extract leaf color distribution. All features are extracted from digital leaf image and the features can be extracted automatically. Combination of leaf feature is done using Product Decision Rule (PDR) and classification of medicinal plants is done using Probabilistic Neural Network (PNN).
2012 International Conference on Advanced Computer Science and Information Systems (ICACSIS) 1-2 Des 2012. pp. 255-259.
II. LOCAL BINARY PATTERN (LBP) Local Binary Pattern (LBP) is proposed by [13] for rotation invariant texture classification. To obtain LBP value, thresholding performed on the neighborhood circular pixels using the central pixel and then multiply by binary weighting. LBP can be formulated as: −1 ,
( ,
)=
(
− )2
(1)
=0
( )= 1 0
≥0