Geographic object-based image analysis (GEOBIA) of Landsat 8 OLI for landform identification

Geographic Object-Based Image Analysis (GEOBIA) is an emerging approach in remote sensing image analysis and classification which relies on segments or objects created by a group of pixels on the image. GEOBIA has been utilized for many remote sensing applications with various degree of success. However, from the literature, its application for landform analysis and classification is still rare. This study aims to test GEOBIA interpretation capabilities to identify landform in part of Opak Watershed (Central Java, Indonesia) using Landsat 8 OLI and DEMNAS imagery (30 and 8- meters pixel size, respectively) and evaluate the result. Both image data were fused to create an image with high spectral and spatial resolution and contains elevation data, as an input for the segmentation process. GEOBIA interpretation process was performed gradually; first, initial Multiresolution Segmentation Algorithm was conducted to identify the variation of slope found in the study site. Then, the slope segments/objects were used to identify landform using Ruleset-Based Classification considering the image object information including object values, pattern, shape, and other parameters. The accuracy of the result was evaluated based on the percentage accuracy of the landform classification. From this study, we found that fusion-image and GEOBIA are capable of distinguishing landform elements very well with the percentage of overall accuracy is 88. This result shows that GEOBIA has potential in identifying and classifying landform objects. © 2019 SPIE.