KLASIFIKASI TUTUPAN LAHAN BERBASIS OBJEK PADA CITRA SENTINEL-2A DI WILAYAH SEMPADAN DANAU POSO

Abstract

Lake Poso is one of the lakes that is included in the 15 critical lakes. Critical lakes are lakes that have experienced extensive narrowing, environmental degradation and are overgrown with water hyacinth. The purpose of this study was to determine object-based land cover using Sentinel 2A imagery on Softwae eCognition 9.01 in the Lake Poso border area. Lake border line regulations are regulated in the Minister of Public Works and Public Housing (PURP) Regulation No. 28/PRT/M/2015 concerning Stipulation of River Border Lines and Lake Border Lines. The research method used is object-based classification or image segmentation using eCognition 9.01 software and Arcgis 10.2.2 software. In identifying land cover, a multi-resolution image segmentation process is carried out for sampling cover classes. The parameters used in retrieval of image objects are scale parameter 100, shape 0.2 and compactness 0.5 to determine land cover objects. Based on the results of the object-based land cover classification analysis in the Lake Poso border area, there are six land cover classes, namely: settlements 152.05 ha, swamps 359.81 ha, secondary forest 23,657 ha, shrubs 254.05 ha, open land 213.8 ha and clouds with an area of ​​209.19 Ha. In the calculation of the accuracy test, there are 30 coordinate points using the confusion matrix formula and get the average of the results of the overall accuracy test for the Lake Poso border area of ​​80%.