Sri Atmaja, Novizar Nazir, Adhi Harmoko Saputro,Rahmat Hidayat, Ario Betha Juanssilfero, Slamet Riyadi

International Journal on.Advanced Science, Engineeringand Information Technology

Vol. 3 (2013) No.1ISSN: 2088-5334

INSIGHT Pub.

n _-\ aja_ ovizar 1\azirAdhi Hannoko SaputroRahmat HidayatArio Betha JuanssilferoSlamet Riyadi

International Journal on Advanced Science,Engineering and Information TechnologyVol. 3 (2013) No.1

This work is subject to copyright. All rights are reserved, whatever the whole or part ofmaterial is concerned, specially the rights of translation, reprinting, re-use of illustratirecitation, broadcasting, reproduction on microfilms or in other way and storage in data baDuplication of this publication or parts thereof is permitted only under the permission 0

publisher.

ISSN: 2088-5334

http://ijasei t. insightsoci ety. org

Copyright © 2012 by INSIGHT PublishingPrinted in Indonesia

11

Preface

.: :S o1r great pleasure to present the Vol. 3 of the International Journal on Advanced Science,engineering and Information Technology (IJASEIT). This volume comprises articles whichare presented in the 1st International Conference Sustainable Agriculture, Food and Energy(SAFE20 13) held in Padang, West Sumatera, Indonesia, 12-13 May 2013. Article submissionscame from different countries that cover varies topics in science, engineering and sustainability.This volume consists of 119 articles classified into six issues based on their field of study.

We would like to take this opportunity to thank all colleagues who had submitted their articlesto the IJASEIT through the committee of the 1st SAFE2013. A lot of number of submissionsindicates their high trustworthy to us to publish their current findings and spread to wideacademic communities. We also send our appreciation to all reviewers who had dedicated theirvaluable time and comments to ensure articles significantly contribute to science andtechnology. In addition, we would like to acknowledge the organizing committee of the151 SAFE 2013 for this great collaboration and the Editorial Board who had worked hard toprepare this volume.

\ .e are pleased to inform you that the editorial boards of the journal have been trying to widenthe joumal indexing to main databases and to receive regular submissions for publication forthe forthcoming issues. We are committed to serve a fast publication and provide quick accessto the recent articles for academic communities globally.

Finally, we do hope that articles published in this volume might inspire a state of the artresearch and new findings for the advancement of science, engineering and informationtechnology.

May 2013Sri Atmaja

Novizar NazirAdhi Harmoko Saputro

Ralunat HidayatArio Betha luanssilfero

Slamet Riyadi

III

Ta e of Content

"l'owards Sustainable Green Pr=-reshFruit Bunches (FFB L-·

Muhammad "'f. ','-

•••.non: Exploring Automated Grading for Oil Palmine Vision and Spectral Analysis 1

Changes of Eco-Friendly Hot Oil Treatment on'n'f-rn-",.T"....-i ,"-•.•••-<. •••• Hybrid 6

101/, Izyan Khalid, Mohd Sukhairi Mat Rasat & Mahmud Sudin

•••.__I•.••porting Sustainable Mangosteen (Garcinia mangos tan a L.) Production ..... 161/ Muas, M. Jawal A,S, and Affandi

. ed System for Tropical Leaf Medicinal Plant Identification 23_ i. Elvira Nurfadhilah, Ervizal A.M. Zuhud, Ellyn K. Damayanti, K.i Arai, H Okumura

sement of Occupational Health and Safety in Rice Farmers in Ngrendeng,- J va in 2012 28Anisa Yonelia, L. Meily Kurniawidjaja

ning Rural Area for the Development of Annual Plants 33'Bariot Hafif , Junita Barus and Masganti

illen and Anther Cultures as Potential Means in Production of Haploid Kenafrib' ells Cannabinus L.) 38if .7IFJ, Ahmed Mahmood Ibrahim And Zed Ermiena Surya Mat Hussin

rial Use of Entomopathogenic Virus Native to Sumatra Island as Biological Controlt of Setora nitens L. (Lepidoptera.Llmacodidae), The Main Pest of Oil Palm 41

::.parman, Yulia Pujiastuti, Hisanori Banda, Asano Shin-Ichiro

dy on Bacillus thuringiensis Indigenous Highland of South Sumatera-Basedninsecticide Towards Lepidopteran Insect Pests 46Yulia Pujiastuti, A. Muslim, Hisanori Bando, Shin-Ichiro Asano

ort-Term Effect of China Violet Compost on Soil Properties of Ultisol and Peanut Yield 50Heri Junedi, Zurhalena, Itang Ahmad Mahbub

lizing Palm Oil Mill Effluent Compost for Improvement of Acid Mineral Soil Chemicaliperties and Soybean Yield 54Ermadani, Arsyad AR

of Extension Agents Towards Agricultural Practice in Malaysia 59"_ • Jill ArifShah, Azizan Astnuni. Azahari Ismail

v

Relation hip Between Decision Making Inputs and Productivity Among Paddy Farmers'in Integrated Agriculture Development Area (IADAs), in Malaysia 64-

. 'IIr Bahiah Mohamed Haris, Azimi Hamzah, Steven Eric Krauss, ISlIIi Arif Ismail

Construction of Hevea Brasiliensis Genetic Linkage Map and Identification ofQuantitative Trait Loci Using Rapd Markers 71""'"

Novalina and Aidi Daslin Sagala

Potential Endophytic Bacteria for Increasing Paddy Val' Rojolele Productivity 71Desriani, Dwi Endah Kusumawati, Akhmad Rivai, Neneng Hasanah, Wiwit Amrinola,Lita Triratna and Ade Sukma

Quality Control of Mass Rearing of Egg Parasitoids of Yellow Rice Stem BorerScirpophaga Incertulas Walker 7

Wilyus, Fuad Nurdiansuah, Asni Johari, Siti Herlinda, Chandra Irsan, Yulia.Pujiastuti

Analysis of Agricultural Rice Production in Jambi: Study of Acreage ResponseUnder Price Policy Program S .

~M -

Pico Hydropower Application on Tidal Irrigation Canal Supporting the IndonesianAgricultural Activities Case Study: Telang Ii-Banyuasin ft;~;

Darmawi, Riman Sipahutar, Siti Masreah Bernas, Momon Sodik Imanuddin.

Prediction of Erosion Rate at Several Land Units in Upper Watershed ofBatang Mangau, 'Vest Sumatra .

Amrizal Saidi, Adrinal and Anggi Kharisma

VI

ournalonSciencee r I n 9echnology

Vol.3 (2013) No.1ISSN: 2088-5334

_4.Computer Aided System for Tropical Leaf MedicinalPlant Identification

Yeni Herdiyeni'", Elvira Nurfadhilah', Ervizal A.M. Zuhud2, Ellyn K. Damayanti",

Kohei Arai ', Hiroshi Okumura3

IDepartment of Computer Science, Faculty of Mathematics and Natural SciencesBogar Agricultural University, West Java, Indonesia

E-mail: #yel1i.herdiyel1i@ipb.ac.id

I~

2Departll1el1tofForest Resources Conservation and Ecotourism, Faculty of Forestry

Bogar Agricultural University, West Java, Indonesia

3Graduate School of Science and Engineering,Saga University, Saga City, Japan;<J:

eX

e objective of this paper is to develop a computer aided system for leaf medicinal plant identification using Probabilisticrk, In Indonesia only 20-22% of medicinal plants have been cultivated. Generally, identification process of medicinaldone manually by a herbarium taxonomist using guidebook of taxonomy/dendrology. This system is designed to help

. entify leaf medicinal plant automatically using a computer-aided system. This system uses three features of leaf toicinal plant, i.e., morphology, shape, and texture. Leaf is used in this system for identification because easily to find.icinal plant we used Probabilistic Neural Network. The features will be combined using Product Decision Rule (PDR).tested on 30 species medicinal plant from Garden of Biopharrnaca Research Center and Greenhouse Center of Ex-

[.~~!.!·~D of Medicinal Indonesian Tropical Forest Plants, Faculty of Forestry, Bogor Agriculture University, Indonesia.ults showed that the accuracy of medicinal plant identification using combination of leaf features increase untilmparative analysis of leaf features has been performed statistically. It showed that shape is a dominant features for

••••.."---""!oil·on.This system is very promising to help people identify medicinal plant automatically and for conservation andIcinal plants. "

. entification; leaf morphology; leaf shape; product decision rule; medicinal plants; probabilistic neural network.

I. INTRODUCTION

country with a mega biodiversity. For_ Indonesia has more than 38,000 plants

- _~1{)1 the Laboratory of Plant Conservation,- eszryBogar Agricultural University (lPB) has

rropicaI medicinal plant species fromecosystems [2]. Unfortunately only 20-22%"- are cultivated by people. One of the_:.ilization of medicinal plants using the

medicinal plants identification

_ .:1 ation is not easy and required an:- nd with significant experience. Also-_-:r who can identify medicinal plants is-.searcher must bring a book dictionary of

plant during plant identification in the field. This caused byutilization of medicinal plants by community is very low.Therefore, we need a computer-based automatically systemas a tool to help people identify these various types of themedicinal plants.

There were some researches on plant identification.Fourier descriptor was used to extract leaf shape ofDicotyledonous leaf class [4]. The accuracy of system is31. 75%. Basic characteristic of leaf morphological and itsderivatives have been analyzed by [5]. The accuracy ofsystem is 27.22%. The Local Binary Pattern (LBP) is used toextract texture of house plants [6]. The accuracy of system is73.33%.

In this research, we develop a computer aided system forIndonesian leaf medicinal plant identification using three

features, i.e., morphological, texture, and shape and also useProbabilistic Neural Network (PNN) for leaf identification.

II. FEATURE EXTRACTION

A. Morphologicol Feature ExtractionMorphological feature consists of two features, basic and

derivate. The basic feature used in this research wasdiameter, area and perimeter/leaf circumference. Three basic:~a ures can be combined to get eight derivate features likesmooth factor, shape factor, ratio perimeter and diameter,also five features of leaf vein [4].

B. Texture Feature ExtractionLocal Binary pattern was originally designed for texture

description. The operator assigns a label to every pixel of animage using the 3x3-neighbourhood of eaeh pixel with thecenter pixel value and considering the results as binarynumber. The histogram of the label is used as texturedescriptor. Fig. I show illustration of the basic LBP operator.

Binary: 1iooior 1Deciutnl: :!03

Fig. I Illustration of the basic LBP Operator

To be able to deal with textures at different scales, theLBP operator was later extended to use different sizes.Defining the local neighbourhood as a set of sampling pointsevenly spaced on circle centered at the pixel to be labelledallows any radius and number of sampling points. In thisresearch the notation (P,R) will be used for pixelneighbourhoods which means P sampling points on a circleof radius R. Fig. 2 show an example of circularneighbourhoods.

IImB+, , , ..

• 0 ~ .' Q ~• .• \ I... ~

Fig. 2 Circular ncighbourhoods with (8,1), (16,2) and (8,2).

Generally, texture can be characterized by a spatialstructure (e.g. a pattern such as LBP) and the contrast (e.g.Rotation Invariant Variance Measure (V AR), the variance oflocal image texture). Spatial structures vary with rotationwhile contrast does not. Ojala et.al [7J proposed using thejoint histogram of the two complementary features, namelyLBPIV AR, for rotation invariant texture classification. Thedrawback of this approach is that the value of VAR iscontinuous so that a quantization step is needed to calculatethe histogram. LBPV is a simplified but efficient joint LBPand contrast distribution method. In this research we usedLBPV with different operator.

C. Shape Feature ExtractionTo detect edge of leaf, this research used image

thrcsholding and canny edge detector. Boundary of an objectis detected using 4-connccted neighbouring. Point by point

of boundary was traversed and represented a:number, x + iY' Sequence of N complex

zen) = :rn + iYn was transformed using Fourier -

resulting Zen). Z(n)are called as Fourier descript

Shape feature of Fourier descriptors result is ebe invariant with the rotation, dilatation, and transthat purpose, some adj ustments are needed 0

descriptors as result of the Fourier trans'Translation is only affected to the first Fourier di(Z(O)), so the part that was took for the shape desonly the Zen) line with I ::;n ::; N. Rotation eff .affected Fourier descriptors line. Dilatation e;affected with same quantity for every Fourier d.

To remove scaling factor we made division 0

Fourier descriptors with one of family at Zen)example Z(I).

D. Probabilistic Neural Network (PNN)PNN is an Artificial Neural Network (ANN)

radial basis function (REF). REF is a function thatlike a bell that scaling nonlinear variable [8].composed of only three layers: the input layer, tblayer and the summation layers. The main advausing the architecture of PNN training data is easyfast. Weight are 110t "trained" but assigned. Existiwill never be alternated but only new vectors arinto weight matrices when training. So it can be ustime. In this research classifier combination isincrease efficiency of decision making.combination used a combination of two or moreindividual characteristics of decision rules. Rclassification of each feature are produceprobability and posterior probability. Basedprobabilities, classifier combination techniques theused is product decision rule (PDR), sum deci.(SDR), the maximum decision rule (MDR) and thevote rule (MVR) [9].

I CB5ttn.gdam corr6'ctly claSSified IaCcta-acy = xL cQsttng dara

E. EvaluationTypically, performance system IS mainly eva

accuracy of classification.

III. EXPERIMENTAL RESULT

A. Experiment DataThe entire document should be in Times New F

Times font. Type 3 fonts must not be used. Othermay be used if needed for special purposes. Recorfont sizes are shown in Table I.

The system was tested on 30 species medicinal pBiof'armaka garden and at Ex-situ Central ConserTropical Forest medicinal plants indonesia, ForestrBogor Agricultural University (IPB), West Java, IThe number of leaves is 1,440 images. Each \{f claof 48 images with resolution is 270 x 240 pixel.that we use in this re .earch are PandanPandanusaniaryllifolius Roxb (class I), Jar

24

curcas Linn (class 2), Dandang Gcndis /iiutsnutans Lindau (class 3), Lavende r/.. iafficinalis Chaix (class 4), Akar Kuning/

. _'isiaflava L (class 5), Daruju/ Acanthus ilicifolius L. Pegagan/ Centellaasiatica, (Linn) Urban. (class 7),

== Cetttellaasiatica.tl.inn) Urban. (class 8), kemangi/tbasilicum (class 9), IIeri Coleus scutellarioides,3ell/h(class 10), jeruknipis/ Citrus aurantifolia,

. _ (class II), Bidani/ Quisqualislndica L. (class 12),;Cina/ Smilax china (class 13), Tabat Barito/cloidea L. (class 14), NandangGendisKuning (classBungaTelang/ Clitoriaternatea L. (class 16),okan/ NothopanaxscutellariumMerr. (class 17),

• -aI Talinumpaniculatum (jacq.) Gaertn. (class 18),utan/ Urenalobata L. (class 19), SosorBebekihoepinnata/Lam.H'ers (class 20), Nanas Kerang/discolor (L.Her.) Hance(class 21), Seligi/

thusbuxifolius Muell (class 22), RemakDaging/'a bicolor Hassk (class 23), Kumis-Kucing/

tphonaristatus (BI) Miq (class 24), Kemuning/apaniculata [L.. } Jack. (class 25), CincauHitam/

> ;pa/llsfris(class 26), SambangDarah/~_. iacochinchinensisl.our. (class 27), Landiklar alupulinal.indl. (class 28), JambuBiji/

iguajava L. (class 29), and Handeuleum/gphyllumptctum (L.) Griffith (class 30).

!d ification of Morphological, Texture, Shape andI. ures Combination Using PNN

~ - research, we conducted four experiments for plant~a..:ation, i.e., based on morphological featurei,s. cation, texture feature classification, shape featurese·.::ation and combination of three features. In training

o~ ·e used 38 pieces of leaves for each class (species)a . ieces of leaves for each class in testing phase. The.Sl

analysis of leaf features has been performedIy.

'ogical Feature Classification

Iu.. experimenral result showed that average accuracy of-phological features only has the average accuracy is

e species get low accuracy. Fig 3 shows accuracyo dentification for each class using morphological

:00%

80%>_ 0'-'0

om

pientry. iIlaels,

. of puru identification based on leaf morphological feature .

ed :hat morphological feature did not work forAccording to observation, it is caused byariaty of species. Figure 4 showed sample ofra

species which has various morphological feature. Fig. 4showed that on class 3 (DandangGendis) train data or testdata images had various sizes with different angle, so thatmorphological feature especially basic leaf features can notbe used for plant identification. For derivate leafcharacteristic, it's quite good in representing thecharacteristic for leaf with the different size but on the sameshoot angle .

Fig. 4. Samples of image class 3 (DandangGendis) Texture

The experimental result showed that average accuracy ofplant identification using LBPV (8,1) is 53%. Figure 5showed accuracy of leaf identifation for each class usingLBPV. Some species can be identified correctly, for exampleclass 9 and 12. But some species still has low accuracycaused by it has different illumination. Fig. 6 show sampleof image (GadungCina) which has different illumination ..

100% +- - ==1! 1

kl-~lill7 10 13 16 19 22 25 28 1

.. __. .cl~ss. . 1

80%

G"GO%coc..340%u<t:20%

0%

1 4

Fig. 5. Accuracy of plant identification based on LBPY (8.1).

Data Training Data Testing

Figure 6. Samples of image class 13 (Gadungf ina).

Shape

The experimental result show that average accuracy ofplant identification using shape feature is 6-1-%. In thisexperiment we used Fourier Descriptor. Fig. 7 show ed the

_ 'erage accuracy for each class using shape features. Ther.umber of species which can be idetified using shape featuremor than species using texture or morphological texture.

1 4 7 10 13 16 19 22 25 28Class

Fig. 7. Accuracy of plant identification based on Fourier Descriptor.

Fig. 8 showed sample of leaves which has low accuracy.The leaves have a various in shape so it is difficult toidentify correctly.

~~

Data Training Data Testing

Fig. 8 Sample of image class 23 (RemakDaging).

C Feature CombinationThe experimental result show that average accuracy of

plant identification using feature combination increased upto 74.67% as shown in Fig. 9. The feature combination hasbetter accuracy compare to single fature. In this experimentwe used feature combination of morphology, texture andhape. In this research we used product decision rule (PDR)o combine these features. Fig. 11 showed the averagea curacy for each class using feature combination.

100%I

80% ~-._- ·1->- I~60~~

ir-' 1-- .•..

::lt: 40%<!

~20%

07~ , . ,r I r I

1 4 7 10 13 16 19 22 25 28ClassL- --"

Fig. 9 Accuracy of plant identification based on feature combination.

Plant identification using feature combination canm rcase the accuracy of identification. It is showed that

different features can be used for plant identificanshow samples of image from class 16 (BungaTelcan be identified correctly using texture or 1110rho:

Data Training Data Testing

Fig. 10 Sample of image from class 16 (BuugaTelang

100%80%60%40%20%0%

·--.-------- ..--·---75~------ ..5·3·lJ---6.4..%-- ..-----~----------Y<-o __ ~ .I-r-20%-----'./--.t- rel-~~~~~=--.,--...,-~.,__r---

>-uco...:Juu<!

Fig. II Comparison of average accuracy using different f

In this research we also performed scomparative analysis of leaf feature combination.that shape is a dominant features in plant idercompare to morphology and texture as shown in Fi,

100%80%60%40%20%0%

~ice.;~o

'S-0o<..~

~L-.... .__. _

Fig. 12 Comparative analysis of leaf features.

IV. CONCLUSIONS

A Computer aided system for tropical leaf mediidentification has been successfully irnplemenidentification is done based on leaf features cc(morphology, shape and texture). The classificausing the PNN method with the PDR classifier ccis quite good to increase the medicine plant idaccuracy up to 74.67%. The comparative featurshow that shape feature is most dominant than rrand texture. This system is very promising to I.

26

_ -. Ian automatically. Also this system can- ervation and utilization of medicinal plants.

ACKNO\VLEDGMENT

rs would like thank to Department of Forest" . d Ecotouri:' ervation an cotounsrn, Faculty of Forestry,

=-:.::ultural University and Department of- Science, Saga University, Japan for research:1. Their support is gratefully acknowledged

REFERENCES

;).

as. 2003. Indonesia Biodiversity Strategy and Action3-2020. Jakarta: Bappenas.

- E.A.M. 2009. Potensi Hutan Tropika IndonesiaPenyangga Bahan Obat AI am untuk KesehatanJumal Bahan Alam Indonesia. Vol. VI No.6, Januari

d. 20 IO. Lokakarya Nasional Tumbuhan Obatesia 2010. Jakarta: Pusat Penelitian dan PengernbanganTanaman.

l--L

...... ..,fea

staI. Ilen?ig

5.

edinerrc

fica'r cidatur11 IT'

:0

27

[4] Herdiyeni, Y., and Ramadhani, M 2010. Shape and VeinExtraction on Leaf Image Using Fourier and B-Spline Model.Asian Federation for Information Technology in Agriculture.Bogor, Indonesia.Annisa. 2009. Morphology and Texture Extraction for LeafRetrieval. Thesis. Department of Computer Science. BogorAgricultural University.Kulsum, Lies U. 20 IO. House Plant IdentificationAutomatically using Local Binary Patterns Descriptor andProbabilistic Neural Network. Thesis. Department ofComputer Science. Bogor Agricultural University.Ojala T., et al. 2002. Multiresolution Gray-Scale and RotationInvariant Texture Classification with Local Binary Pattern.IEEE Transactions on PAMI. Vol. 24, No.7, pp. 2037-2041.Wu S. G., et al. 2007. A Leaf Recognitian Algorithm forPlant Classification Using Probabilistic Neural Network.China: Chinese Academy of Science.Kittler, 1., et al. 1998. On Combining Classifiers. IEEETransactions On Pattern Analysis And Mechine Intelligence.Vol 20 no.3 page: 226-239.

[5]

[6]

[7]

[8]

[9]