COMPARISON OF LAND COVER CLASSIFICATION ACCURACY TEST USING PIXEL BASED AND OBJECT BASED IMAGE ANALYSIST (OBIA) METHODS ON SENTINEL-2A IMAGE IN 2023 IN PADANG CITY

Abstract

Land cover is the physical and biological cover of the earth's surface, whether formed naturally such as swamps, rivers, hills or man-made such as rice fields, gardens, forests and buildings. One alternative to obtain information about land cover is by utilizing remote sensing technology. These include pixel-based classification and object-based classification. The fundamental difference between object-based classification and pixel-based classification lies in the object separation process, object-based classification divides based on segmentation results, not just based on single pixel values ​​in pixel-based classification. This research aims to compare the level of classification accuracy of pixel-based methods with object-based methods in identifying land cover in 2023 Sentinel-2A image data in Padang City. The sampling method used was random sampling with a total of 7 classes. To get accuracy results, the same ground truth data is given to both methods.The results of the comparison of object-based classification and pixel-based classification on the two images were tested for accuracy using a confusion matrix which resulted in land cover classification accuracy on the 2023 Sentinel-2A image. The overall kappa accuracy for the pixel-based classification method was 89.80%, while for the pixel-based classification method the object obtained a value of 94.88%. The overall accuracy results show that object-based classification is better than pixel-based classification in classifying land cover.