An unsupervised framework to extract the diverse building from the satellite images using Grab-cut method

Building detection from the satellite image is a computer vision, Photogrammetry, and remote sensing task that has significant importance in geographical information system (GIS) based applications. In this study, a novel framework is developed for the automatic detection of different types of buildings in the complex environment of the satellite images. The framework consists of fuzzy-based pre-segmentation, information extraction, and Grab-cut partitioning. The pre-segmentation and information extraction are employed to generate the initialization data for the Grab-cut method in an unsupervised manner. Further, the Grab-cut method partition the input image in building and non-building classes depending on the initialization data is provided. The performance of the proposed algorithm is evaluated over pan-sharpened satellite images having diverse built-up characteristics. The qualitative and quantitative assessments are conducted using standard statistical parameters. The proposed algorithm has achieved the average performance in terms of the F - score is 65%, and in terms of recall is 84%. Also, the proposed algorithm is compared with the existing state-of-the-art methods to illustrate the superiority and potential of the proposed algorithm over the existing ones.