Xiaojun YANG, Florida State University, United States
Feilin LAI, Iowa State University, United States
Fang ZHANG, Hainan University, China
Ting LIU, Northeastern Illinois University, United States
Given the importance of land cover information in global change and sustainability science (see Turner et al. 2007), there have been numerous research efforts towards improving thematic mapping accuracy mostly through the use of remote sensing and machine learning technologies. While comprehensive reviews on the status of remote sensing-based land cover mapping were given elsewhere (such as Gómez et al. 2016; Wulder et al. 2018), we herewith direct the attention on some issues and opportunities that could help address the challenges in mapping complex environments characterized by the presence of spectrally and spatially heterogeneous landscape types, such as urban areas. Based on extensive literature review and our own work (e.g., Zhang and Yang 2020; Lai and Yang 2020), we specifically examined several critical areas, which include the use of time-series images and feature selection techniques, advanced machine learning techniques, integration of remote sensor and other types of geospatial data (including social sensed data), and the uncertainty of mapping accuracy. We conclude that successful use of remote sensing for land cover mapping over heterogeneous landscapes is dependent upon an adequate understanding of the nature of landscape complexity, sensor systems, and information extraction techniques employed in relation to specific mapping objectives.
Mots clés : Land cover mapping|remote sensing|complex environments|data integration|machine learning
A102872XY