文献类型：期刊文献

英文题名：Detecting and monitoring urban project development from space: an unsupervised learning model based on InSAR coherence time series

作者：Mohamadi, Bahaa[1];Balz, Timo[2];Pirasteh, Saied[1,4];Li, Huxiong[1];Younes, Ali[3]

机构：[1]Shaoxing Univ, Sch Intelligent Engn, Inst Artificial Intelligence, Shaoxing, Peoples R China;[2]Wuhan Univ, State Key Lab Informat Engn Surveying Mapping & Re, Wuhan, Peoples R China;[3]Kafrelsheikh Univ, Fac Arts, Geog & GIS Dept, Kafrelsheikh, Egypt;[4]Saveetha Inst Med & Tech Sci Chennai, Saveetha Sch Engn, Dept Geotech & Geomat, Chennai, Tamil Nadu, India

年份：2025

外文期刊名：GEO-SPATIAL INFORMATION SCIENCE

收录：SCI-EXPANDED(收录号：WOS:001560334200001)、、EI(收录号：20253619109160)、Scopus(收录号：2-s2.0-105014889089)、WOS

基金：This work is supported by the National Natural Science Foundation of China [Grant number 42250610212].

语种：英文

外文关键词：Coherence change detection (CCD); coherence time series; Greater Cairo; InSAR; machine learning

外文摘要：Remote monitoring of large development projects using optical imagery is always challenging due to the spatial and temporal characteristics of the imagery and the spectral nature of detected projects. Hence, most urban projects are typically observed only after they have already undergone development. This study introduces a SAR coherence-based model that allows monitoring ongoing development projects using long time series. This model can approximately define when the project has been started and/or finished, periods of intensive work, and slow progress during the investigation period. The model relies on Sentinel-1 imagery to detect areas of development projects by classifying InSAR coherence over long time series using unsupervised learning. The model starts with interferometric coherence estimation, followed by a two-step unsupervised learning method: principal component analysis (PCA) and ISODATA clustering algorithms. The interpretation of the unsupervised classes was done using statistical trends of coherence and long-time series information. The classified map was interpreted into five main classes: stable, seasonal trend, positive first-order trend, negative first-order trend, and negative second-order trend. This model was tested on Greater Cairo, Egypt, by analyzing three years of monthly InSAR coherence time series, and the accuracy of the unsupervised classification map in the tested area was 92.7%. To validate the proposed model, information was gathered from 39 development projects in Greater Cairo, and the model's efficacy in detecting these projects was evaluated. The validation demonstrated that the model was able to detect changes in 38. The model shows a strong ability to detect building construction projects, building demolition-induced new-axis roads, and most bridge constructions. This study extended the importance of InSAR coherence to a new level by utilizing long time series information for detailed monitoring of urban project development from space. With the global coverage advantage of Sentinel-1 imagery, this model is applicable for monitoring development projects in any city worldwide.