Machine learning approach to extract building footprint from high-resolution images: the case study of Makkah, Saudi Arabia

Abstract

Abstract Extracting and identifying building boundaries from high-resolution images have been a hot topic in the field of remote sensing for years. Various methods including geometric, radiometric, object based and edge detection were previously deliberated and implemented in different studies in the context of building extraction. Nevertheless, the reliability of extraction process is mainly subject to user intervention. The current study proposes a new automatic morphology-based approach for extracting buildings using high-resolution satellite images of Al-Hudaybiyah region in the city of Makkah as a case study. The proposed technique integrates the support vector machine for extracting buildings that have bright and dark roofs. The appropriateness of this method has been examined by means of various indicators for example completeness, correctness and quality. Preliminary findings will illustrate the precision and accuracy of the used machine learning algorithm. Research results can provide a generic indicator to assist the planning authorities in achieving better urban planning processes taking into account all potential environmental, social and urban demands and requirements.