COMPARISON OF PIXEL-BASED AND OBJECT-BASED LAND USE LAND COVER CLASSIFICATION USING SATELLITE IMAGERY

Prawira, Kadek Losinanda (2025) COMPARISON OF PIXEL-BASED AND OBJECT-BASED LAND USE LAND COVER CLASSIFICATION USING SATELLITE IMAGERY. Undergraduate thesis, Universitas Pendidikan Ganesha.

[img] Text (COVER)
2115101046-COVER.pdf
Download (825kB)
[img] Text (ABSTRAK)
2115101046-ABSTRAK.pdf
Download (168kB)
[img] Text (BAB 1 PENDAHULUAN)
2115101046-BAB 1 PENDAHULUAN.pdf
Download (150kB)
[img] Text (BAB 2 KAJIAN TEORI)
2115101046-BAB 2 KAJIAN TEORI.pdf
Restricted to Repository staff only
Download (412kB) | Request a copy
[img] Text (BAB 3 METODELOGI PENELITIAN)
2115101046-BAB 3 METODELOGI PENELITIAN.pdf
Restricted to Repository staff only
Download (556kB) | Request a copy
[img] Text (BAB 4 HASIL DAN PEMBAHASAN)
2115101046-BAB 4 HASIL DAN PEMBAHASAN.pdf
Restricted to Repository staff only
Download (2MB) | Request a copy
[img] Text (BAB 5 PENUTUP)
2115101046-BAB 5 PENUTUP.pdf
Restricted to Repository staff only
Download (126kB) | Request a copy
[img] Text (DAFTAR PUSTAKA)
2115101046-DAFTAR PUSTAKA.pdf
Download (213kB)
[img] Text (LAMPIRAN)
2115101046-LAMPIRAN.pdf
Download (170kB)

Abstract

The increasing global demand for land use has led to significant changes in the environment. Land use and land cover (LULC) classification is an important part of environmental control as it provides information on land use changes and patterns over time. This study aims to develop a LULC classification model for Chonburi Province, Thailand using pixel-based and object-based classification approaches in the Google Earth Engine Platform, and to evaluate their performance comparatively. The classes to be classified include Bare soil, Built-up, Forests, Agriculture, and Water. A total of 500 data were used with details of each class consisting of 100 data. The data is divided into 70% for the training set and 30% for the testing set. The main stages in model development include data composition, feature extraction, reference data labelling, classification, and performance evaluation. The classification model uses the Random Forest (RF) and Support Vector Machine (SVM) algorithms. In object-based classification, image segmentation is an additional step performed before classification which is done using the Simple Non-Iterative Clustering (SNIC) and G-Means algorithms. Performance evaluation is carried out using a confusion matrix with accuracy, recall, precision, and F1-score metrics along with image testing. The results show that the object-based method produces higher accuracy compared to the pixel-based one. In pixel-based classification, both RF and SVM models achieved an accuracy of 92%. In object-based classification, the combination of G-Means with RF, GMeans with SVM, SNIC with RF, and SNIC with SVM models achieved an accuracy of up to 93.34%. In addition, the visualization of the object-based classification results is also neater and better organized. The SNIC with RF model produces the best visualization among all tested approaches. The results of this study are expected to contribute in the form of a more optimal and efficient method for monitoring LULC changes.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Remote Sensing, Sentinel-2, Machine Learning, Pixel-based LULC Classification, Object-based LULC Classification
Subjects: T Technology > T Technology (General)
Divisions: Fakultas Teknik dan Kejuruan > Jurusan Teknik Informatika > Program Studi Ilmu Komputer (S1)
Depositing User: Kadek Losinanda Prawira
Date Deposited: 30 Jan 2025 04:28
Last Modified: 30 Jan 2025 04:28
URI: http://repo.undiksha.ac.id/id/eprint/22930

Actions (login required)

View Item View Item