AFDE-Net Building Change Detection Using Attention-Based Feature Differential Enhancement for Satellite Imagery
Author(s):
katroth Balakrishna Maruthiram, Dr.G.Venkata Rami Reddy, Munigala Anusha
Keywords:
Buildings, Feature extraction, Satellites, Semantics, Remote sensing, Computer architecture, Change detection algorithms
Abstract
Agricultural land the board, building change detection (BCD) utilizing satellite pictures, and GIS information base updates all rely upon it. Notwithstanding, deep learning-based change discovery strategies, which often focus on variety and surface, experience issues with regards to perplexing varieties in building rooftops that copy their environmental elements. Moreover, loss of spatial data from down testing could bring about lopsided result limits and incomplete developments. We present AFDE-Net, a remarkable Siamese organization that joins consideration modules and differential picture qualities with a learnable boundary, to defeat these issues. To decrease the deficiency of spatial data and improve the nature of profound elements in high-layered inputs, AFDE-Net uses anensemble spatial-channel attention fusion (ESCAF) module related to a deep supervision (DS) module. We likewise give EGY-BCD, a novel dataset made to recognize building changes that comprises of high-goal, multitemporal satellite pictures from four Egyptian waterfront and metropolitan areas. Profound learning calculations are tested by EGY-BCD on the grounds that it contains photographs with muddled types of progress, like thick structures with rooftops that imitate their environmental elements. On the EGY-BCD dataset, AFDE-Net performs better compared to different methodologies, with a overall accuracy (OA) of 94.3%, a F1-score of 88.8%, and a mIoU of 86.6%.
Article Details
Unique Paper ID: 167132

Publication Volume & Issue: Volume 11, Issue 3

Page(s): 279 - 285
Article Preview & Download


Share This Article

Join our RMS



Conference Alert

NCSEM 2024

National Conference on Sustainable Engineering and Management - 2024

Last Date: 15th March 2024

Latest Publication

Call For Paper

Volume 11 Issue 1

Last Date for paper submitting for Latest Issue is 25 June 2024

About Us

IJIRT.org enables door in research by providing high quality research articles in open access market.

Send us any query related to your research on [email protected]

Social Media

Google Verified Reviews

Contact Details

Telephone:6351679790
Email: [email protected]
Website: ijirt.org

Policies

Important Links

Browse Rsearch papers