Analysis of object oriented classification on high resolution images

Abstract

Classification is one step of image processing which aims to obtain information from remote sensing data. Object oriented classification is a classification algorithm which can be used to process a high resolution image because it uses the object elements as spectral, spatial, and texture. The purpose of this study is to examine the application of object oriented classification on two types of images which are categorized as high resolution images. The study uses two types high resolution image. They are SPOT 5 and QuickBird image covering Ngaglik district, Sleman, Yogyakarta. SPOT 5 was acquired on September 2, 2014 which has 2.5 meters spatial resolution and QuickBird was acquired in 2010 which has 0.6 meters spatial resolution. A field survey was conducted for identifying the classification scheme and testing the classification accuracy. Classification has been done through several stages, involving segmentation, merge parameter determination, rule-based classification, and post classification. Object oriented classification extracts meaningful image objects into several classes based on their spectral, spatial, and texture elements. The utilization of object's spatial element is the advantage of object oriented classification in making classification on the high resolution image. The result of this study shows that object oriented classification algorithm can classify the land use on SPOT 5 image in up to 80.00% accuracy with 16 classes, and QuickBird image in up to 92.00% accuracy with 17 classes. Therefore, object oriented classification is very effective to make classification on the high resolution image. Object oriented classification which uses the high resolution image can produce a good accuracy.