Integrating object-based and pixel-based segmentation for building footprint extraction from satellite images

• DOI: 10.1016/j.jksuci.2023.101802
• Publication Date: 2023-10-19T04:08:54Z
• Authors: Sohaib K.M. Abuja...; Rania Almajalid; Raniyah Wazirali; Rami Ahmad; Enes Taşoğlu; Ismail R. Karas; Ihab Hijazi
• Keywords: Footprint

Accurately delineating building footprints from optical satellite imagery presents a formidable challenge, particularly in urban settings characterized by intricate and diverse structures. Consequently, enhancing the utility of these images for geospatial data updates demands meticulous refinement. Machine learning algorithms have made notable contributions this context, yet pursuit precision remains an ongoing challenge. This paper aims to enhance accuracy footprint extraction through integration object-based pixel-based segmentation techniques. Additionally, it evaluates performance machine methodologies, specifically LightGBM, XGBoost, Neural Network (NN) approaches. The model's evaluation employed low spectral resolution images, widely accessible cost-effective acquisition. study's outcomes demonstrate substantial enhancement compared extant literature. proposed methodology attains overall 99.39%, F1 measurement 0.9935, Cohen Kappa index 0.9870. Thus, approach signifies noteworthy advancement over existing techniques high-resolution imagery.