Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/72535
Title: Kajian beberapa metode klasifikasi citra digital terhadap data penginderaan jauh
Other Titles: Study of several digital image classification methods for remote sensing Data
Authors: Wijayanto, Hari
Wiweka
Muhammad, Faizal Teguh
Issue Date: 2014
Publisher: Bogor Agricultural University (IPB)
Abstract: Klasifikasi citra digital terdiri dari banyak metode alternatif yang menghasilkan tingkat akurasi berbeda-beda. Akurasi ini sangat tergantung pada beberapa hal seperti training sample dan keragaman kenampakan lahan pada daerah citra yang dikaji. Penelitian ini bertujuan untuk membandingkan 4 metode klasifikasi citra digital yang diterapkan pada daerah dengan tingkat keragaman kenampakan lahan yang berbeda yaitu Kecamatan Ciomas, Kecamatan Dramaga dan Kecamatan Cibungbulang. Metode klasifikasi citra digital yang digunakan pada penelitian ini adalah kemungkinan maksimum, jarak Mahalanobis, jaringan syaraf tiruan dan support vector machine. Hasil penelitian ini menyimpulkan bahwa kemungkinan maksimum merupakan metode klasifikasi citra yang paling baik pada citra di tiga kecamatan terpilih dengan nilai rata-rata akurasi keseluruhan sebesar 91.99% dan nilai rata-rata koefisien kappa sebesar 0.8772. Selain itu, metode support vector machine dan jaringan syaraf tiruan juga memberikan hasil yang cukup baik.
Digital image classification is composed of many alternative methods that produced different levels of accuracy. This accuracy depends on several things such as training samples and the diversity of land on the appearance of the image regions studied. This study aimed to compare 4 digital image classification method which applied to areas with different land appearance diversity of the District Ciomas, District Dramaga and District Cibungbulang. Digital image classification methods that used in this study are maximum likelihood, Mahalanobis distance, artificial neural networks and support vector machine. The results of this study concluded that the maximum likelihood is the best image of image classification method of the three district chosen with the average value of overall accuracy was 91.99% and the average value of the kappa coefficient was 0.8772. In addition, support vector machine and artificial neural networks also provide good results. Keywords: commission error, image classification, kappa coefficient, omission error, overall accuracy
URI: http://repository.ipb.ac.id/handle/123456789/72535
Appears in Collections:UT - Statistics and Data Sciences

Files in This Item:
File Description SizeFormat 
G14ftm.pdf
  Restricted Access
Full Text624.15 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.