Back to Search Start Over

Unsupervised Change Detection around Subways Based on SAR Combined Difference Images

Unsupervised Change Detection around Subways Based on SAR Combined Difference Images

Authors :
Aihui Jiang
Jie Dai
Sisi Yu
Baolei Zhang
Qiaoyun Xie
Huanxue Zhang
Source :
Remote Sensing, Vol 14, Iss 17, p 4419 (2022)
Publication Year :
2022
Publisher :
MDPI AG, 2022.

Abstract

Prompt and precise acknowledgement of surface change around subways is of considerable significance in urban rail protection and local environmental management. Research has proven the considerable potential of synthetic aperture radar (SAR) images for detecting such information; however, previous studies have mostly focused on change intensity using single Difference images (DIs), e.g., difference value DI (DVDI) and mean value DI (MVDI). With the aim of more accurate information with respect to surface changes around subways, in this study, we proposed a novel SAR detection method that involved three steps: (1) the calculation of three single DIs, (2) the combination of the single DIs and (3) the delineation of the changed area. Compared to existing detection methods, the proposed method represents three major improvements. First, both the intensity information and phase information were applied by combining the DVDI, MVDI and coherent difference images (CDIs). Secondly, a local energy weight (LEW) approach was proposed to combine single DIs instead of the normally used equal weights. Because the changed area often comprises continuous rather than discrete pixels, a combined DI with the LEW (“CoDI-LEW” hereafter) fully considers the attributes of adjacent pixels and enhances the signal-to-noise ratio of SAR images. Thirdly, the FCM algorithm, instead of the widely used threshold methods, was applied to distinguish changed areas from unchanged areas. An experimental comparison with several existing detection methods showed that the proposed method could delineate changed areas with higher accuracy in terms of both quality and quantity. Furthermore, it can effectively execute detection under diverse surface change conditions with good feasibility and applicability.

Details

Language :
English
ISSN :
20724292
Volume :
14
Issue :
17
Database :
Directory of Open Access Journals
Journal :
Remote Sensing
Publication Type :
Academic Journal
Accession number :
edsdoj.f69c6167a8c14d07a5d11d8c18999a16
Document Type :
article
Full Text :
https://doi.org/10.3390/rs14174419