Building Footprint Identification Using Remotely Sensed Images: A Compressed Sensing-Based Approach to Support Map Updating DOI Creative Commons
Rizwan Ahmed Ansari, Rakesh Malhotra, Mohammad Zaheer Ansari

и другие.

Geomatics, Год журнала: 2025, Номер 5(1), С. 7 - 7

Опубликована: Янв. 31, 2025

Semantic segmentation of remotely sensed images for building footprint recognition has been extensively researched, and several supervised unsupervised approaches have presented adopted. The capacity to do real-time mapping precise on a significant scale while considering the intrinsic diversity urban landscape in data consequences. This study presents novel approach delineating footprints by utilizing compressed sensing radial basis function technique. At feature extraction stage, small set random features built-up areas is extracted from local image windows. are used train neural network perform classification; thus, learning classification carried out domain. By virtue its ability represent characteristics reduced dimensional space, scheme shows promise being robust face variability inherent images. Through comparison proposed method with numerous state-of-the-art different spatial resolutions clutter, we establish robustness prove viability. Accuracy assessment performed segmented footprints, comparative analysis terms intersection over union, overall accuracy, precision, recall, F1 score. achieved scores 93% 90.4% 91.1% score, even when dealing drastically features. results demonstrate that methodology yields substantial enhancements accuracy decreases dimensionality.

Язык: Английский

Building Footprint Identification Using Remotely Sensed Images: A Compressed Sensing-Based Approach to Support Map Updating DOI Creative Commons
Rizwan Ahmed Ansari, Rakesh Malhotra, Mohammad Zaheer Ansari

и другие.

Geomatics, Год журнала: 2025, Номер 5(1), С. 7 - 7

Опубликована: Янв. 31, 2025

Semantic segmentation of remotely sensed images for building footprint recognition has been extensively researched, and several supervised unsupervised approaches have presented adopted. The capacity to do real-time mapping precise on a significant scale while considering the intrinsic diversity urban landscape in data consequences. This study presents novel approach delineating footprints by utilizing compressed sensing radial basis function technique. At feature extraction stage, small set random features built-up areas is extracted from local image windows. are used train neural network perform classification; thus, learning classification carried out domain. By virtue its ability represent characteristics reduced dimensional space, scheme shows promise being robust face variability inherent images. Through comparison proposed method with numerous state-of-the-art different spatial resolutions clutter, we establish robustness prove viability. Accuracy assessment performed segmented footprints, comparative analysis terms intersection over union, overall accuracy, precision, recall, F1 score. achieved scores 93% 90.4% 91.1% score, even when dealing drastically features. results demonstrate that methodology yields substantial enhancements accuracy decreases dimensionality.

Язык: Английский

Процитировано

0