| Abstrak/Abstract |
Image segmentation algorithms for information extraction are currently growing rapidly. One of the uses of segmentation operations is to delineate building objects. Delineation and inventory of building objects is very important to map the current status of land cover, support urban planning, and calculation of losses due to disasters. However, detection of building objects from remote sensing images using segmentation algorithms is not easy. Various segmentation parameters are influencing the accuracy of segmentation results. This study aims to determine the optimum segmentation parameterization of building objects in urban areas using high-spatial resolution image data. The location of the study sample was in parts of Padang City, West Sumatra, and the image used was GeoEye-1 acquired on January 2018. Various combinations of scale, shape and compactness parameters were simulated to obtain the best segmentation parameters. The reference used to assess segmentation accuracy was the result of visual interpretation of building objects. Five measures of accuracy assessment were used, including over-segmentation (OSeg), under-segmentation (USeg), root mean square error (D), area fit index (AFI), and quality rate (Qr). This study found that shape and compactness parameter have high contribution in recognizing building objects from the image. The optimal parameters for segmenting building object using GeoEye-1 imagery were using a combination of image band weight of 1122, scale parameter of 70, shape parameter of 0.3 and compactness parameters of 0.7. This study reveals the most relevant segmentation parameters to be used in discriminating building from high-spatial resolution image data. |
