#TITLE_ALTERNATIVE#

• Main Author: YUDHISTIRA (NIM : 15108065); Pembimbing : Prof. Ketut Wikantika, Ph.D & Dr. Soni Darmawan,, RAHDHITYA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/19207
• Institution: Institut Teknologi Bandung

A remote sensing imagery can be used for many things nowadays. One of the most popular usages is land cover classification. Classifying the land cover of an area can produce a lot of useful information, such as the development of infrastructure in a city or the diversity of building in a city. Nowadays, there are a lot of methods use to classify land cover of some area using remote sensing imaging. One of the promising new methods is Support Vector Machines (SVMs). SVMs themselves are based from a statistical learning theory, which is, in basic, a binary classifier. In this research, SVMs are used to find out whether the overall accuracy of this method is better compared to another classifying method, Maximum Likelihood (MLC), when used to classify some areas in Bandung Regency using Landsat TM-5 data. The results showed that mainly SVMs fared better than MLC, especially in homogenous area where there are only 2-3 classes in those areas. Although there were some problems for SVMs when dealing with a rather dense area where there are 3 classes or more. Overall, results showed us that SVMs were mainly better than MLC, especially when dealing with homogenous areas. However, MLC were slightly better in dealing with heterogeneous areas, even though SVMs were not far behind. The solution to this could be with applying the kernel function in future studies to gain a better result.